Can anti-corruption improve the quality of environmental information disclosure?

This study takes the anti-corruption campaign of the 18th CPC National Congress in China as an exogenous impact build a double difference (DID) model, proving that anti-corruption can increase the quality of enterprise environmental information disclosure significantly. After a series of tests such as parallel trend test and placebo test, the results are still robust. The intermediary effect model results indicate that the anti-corruption increases the violation cost of enterprises and encourages enterprises to carry out innovation activities, and ultimately increase the quality of enterprise environmental information disclosure. Heterogeneity analysis finds that non-state-owned enterprises, large enterprises, and regions with more substantial environmental supervision will improve the quality of enterprise environmental information disclosure to a greater extent. The findings of this paper can help provide policy inspiration for improving the mandatory environmental information disclosure system and strengthening the enforcement of environmental regulations.


Introduction
As the booming of China's economy, serious environmental pollution has proved to be a key factor restricting society's overall development (Du et al. 2013;Dogan and Seker 2016;Dogan and Ozturk 2017;Wang and Sun 2018). Chinese government has introduced a series of environmental protection regulations in recent years, trying to solve high environmental costs in the process of economic development. As an essential channel for external stakeholders to obtain corporate environmental performance, environmental information disclosure is an integral part of China's modern environmental regulatory system. The improvement of quality of enterprise environmental disclosure is conducive to public's real understanding of corporate environmental performance, forcing enterprises to carry out environmental governance and improving environmental conditions (Wang et al. 2017a).
Up to now, China has issued several regulations to standardize environmental information disclosure (Wang and Zhang 2019), such as the Environmental Information Disclosure (Trial), the Measures for Environmental Information Disclosure of Enterprises and Institutions, etc. These regulations stipulate heavy polluting enterprises, which are the main sources of pollutant emission, to disclose environmental information about pollutant emissions, the operation of environmental protection facilities, environmental input, etc. More importantly, for enterprises that disclose environmental information illegally, the environmental protection departments will pursue the corresponding legal responsibility and impose a fine.
Although above regulations put forward higher requirements on the quantity and quality of environmental information disclosure, existing studies indicate that the quality of enterprise environmental disclosure has not been substantially improved. Enterprises tend to disclose only self-interest information, mainly a qualitative description (Liu and Anbumozhi 2009). In 2019, China Association of Environmental Journalists and Beijing University of Chemical Technology jointly released the Evaluation Report on Environmental The original online version of this article was revised due to a retrospective Open Access cancellation. Information Disclosure of Chinese Listed Companies. The report showed that the average score of the environmental information disclosure index of Chinese listed enterprises was only 33.44 points, which was still low. Thus, improving quality of enterprise environmental disclosure is still an urgent matter that needs to be settled. China's existing environmental regulation system has a serious problem of paper based; there is a large gap between legislations and enforcement. Enterprises can collude with local environmental protection authorities through rentseeking and other ways to adopt strategic interaction, so as to escape the constraint of environmental regulation (Fisman and Svensson 2007). Under such condition, the incentive effect generated by the cost of violation is difficult to be internalized in the production and operation activities of enterprises, and will not promote the quality of enterprise environmental information disclosure (Paunov 2016).
On November 4, 2012, Xi Jinping, General Secretary of the CPC Central Committee, stressed that corruption had seriously damaged the ruling foundation of the Party and declared a firm fight against corruption. After that, a series of anti-corruption regulations have been implemented, including "eight policies," "six prohibitions," etc.; the national anti-corruption efforts have increased unprecedentedly. A large number of government officials and state-owned enterprises' executives have been investigated and punished for corruption, and the corruption of "collapse mode" in some places has also been exposed. According to statistics, from 2013 to 2016, discipline inspection and supervision organs across the country accepted 2.674 million cases of corruption, instituted 1.545 million prosecutions, punished 1.537 million people, and transferred 58,000 people to judicial organs for suspected crimes. In comparison, from 2008 to 2012, only 693,000 people were disciplined by the Party and 27,951 were transferred to judicial organs.
In fact, the anti-corruption has strengthened the punishment of government officials who violate corruption rules, such as overstepping their powers and abusing their power, which made the government's administrative power subject to strict supervision and legal authorization. Such strict regulations on government power and behavior greatly increase the probability of non-productive behaviors being discovered and punished, the cost of non-productive behavior has risen significantly. In addition, the anti-corruption cracks down on government officials' lazy government inaction behavior, and enhances the government execution of environmental policy, so as to improve the intensity of environmental regulation of the government to enterprise (Cai et al. 2011). At this time, enterprises have no choice but to disclose high-quality environmental information as a means to deal with the pressure of environmental regulation (Cheng et al. 2017).
More importantly, anti-corruption clearly standardizes administrative rules and procedures, limits the government's discretion, and reduces the government's excessive intervention in the market (Olken 2007). Therefore, anti-corruption is conducive to the construction of a fair and sound institutional environment, which can effectively guarantee the possibility of enterprises to obtain external resources, and promote fair competition among enterprises. At this time, considering that the net income of rentseeking behavior is limited, enterprises will invest more capital in the field of production, optimize the production structure, improve the energy utilization efficiency, save energy consumption, and reduce pollutant emission (Xu and Yano 2017). The environmental performance of enterprises can be improved, and enterprises are more willing to disclose high-quality environmental information. Above all, there is a close relationship between anticorruption and information disclosure quality, so it is important to discuss the causal relationship between them.
However, most of the existing literatures ignore the influence of macro-institutional level on enterprise environmental information disclosure, and pay more attention to the influence of internal factors on the quality of environmental information disclosure (Clarkson et al.,2008;Lewis et al. 2014;Wang et al. 2017b;Grigoris et al. 2019).This research's marginal contribution is mainly to further explore the impact of external factors on environmental information disclosure from the perspective of anticorruption, in order to make up for the lack of such literature. Based on the practice of Pan and Tian (2017) and Kong et al. (2020), this paper intends to take the anti-corruption campaign of the 18th CPC National Congress in China as an exogenous shock, using the panel data of heavily polluting manufacturing listed companies in Shanghai and Shenzhen A-shares over 2008-2018 to construct the DID model, through a series of robustness tests such as parallel trend test and placebo test, trying to alleviate the endogenous problems in the model.
The structure of rest of this study is as follows: Section 2 is literature review. Section 3 reports the practical design, including data sources and variable definitions. Section 4 shows the estimated results of benchmark regression and robustness test. Section 5 is the mechanism test and heterogeneity analysis. Section 6 gives some conclusions and policy suggestions.

Literature review
There are two types of literatures related to this study, one is the influencing factors of enterprise environmental information disclosure and the other is the micro-impact of anticorruption on enterprises behavior. The above literatures provide guidance for the mechanism analysis and quantitative model setting of this study.

The influencing factors of enterprise environmental information disclosure
The internal factors of the quality of enterprise environmental information disclosure have always been the focus of many scholars. Based on the signal transmission theory, existing studies generally believed that listed companies with good financial performance will disclose high-quality environmental information in order to gain recognition from stakeholders and reduce environmental risks of enterprise sustainable operation (Patten 2002;Zeng et al. 2010;Zhu and Xue 2009). Chen et al. (2020) used the data of 363 listed manufacturing companies from 2012 to 2018 to build GLS model to confirm that there is a positive correlation between financial performance and the quality of environmental information disclosure, and the quality of environmental information disclosure of heavy polluting enterprises is significantly higher than that of non-heavy polluting enterprises. This conclusion is also true for heavy polluting enterprises in the USA (Clarkson et al. 2008). Other studies, based on the principal-agent theory, found that management characteristics such as good management competence and higher educational level of managers also have a positive impact on the quality of environmental information disclosure (Said et al. 2013;Jizi et al. 2014;Wang et al. 2017a;Grigoris et al. 2019).
Some scholars believed that external factors will also affect corporate environmental information disclosure. Considering that environmental problems are typical examples of market failure, which have obvious characteristics of externalism, the existing literatures suggested that the external pressure brought by social supervision and media reports will force enterprises to carry out environmental governance and encourage enterprises to disclose high-quality environmental information (Brammer and Pavelin 2006;Clarkson et al. 2008;Aerts and Cormier 2009). Few scholars have considered the impact of macro-institutional environment on environmental information disclosure. As far as we know, only two scholars have studied the relationship between political association and environmental information disclosure. Charles and Roberts (2006) used the data of 119 heavily polluting enterprises in the USA to construct the Tobit model and confirmed that there is a significant positive correlation between political expenditure and environmental disclosure. Cheng et al. (2017) used the panel data of European and American enterprises for 3-5 years to construct a fixed-effect model to confirm that political connection will reduce the quality of environmental information disclosure, which is a means for enterprises to evade environmental investment and environmental regulation. However, Charles and Roberts (2006) and Cheng et al. (2017) measured the degree of political connection by political expenditures and the number of party members in the company's executive board, respectively, which ignored the endogenous problems caused by the measure bias. The ideal measurement indicators of political connection should not only be independent of anti-corruption efforts, but also have as little relationship with corporate governance behavior as possible, so as to objectively, accurately, and comprehensively reflect the degree of political connection of enterprises.
The micro-impact of anti-corruption on enterprise behavior Many scholars have conducted extensive empirical studies on the micro-impact of anti-corruption on enterprise behavior. Most of these literatures are based on econometric analysis, and the premise of econometric analysis is to measure anticorruption. As far as we know, most scholars measured anticorruption by the number of anti-corruption cases investigated and prosecuted (Glaeser and Saks 2006;Campante and Do 2014). However, this indicator is influenced not only by anti-corruption but also by corruption, which may lead to biased estimation results. Some scholars selected the proportion of articles in the provincial government newspapers of the Communist Party of China (CPC) as the substitute variable for anti-corruption intensity (Xu and Yano 2017). However, this kind of measurement method has strong subjectivity, which may lead to the emergence of systematic bias.
Based on the measurement of anti-corruption, existing research mainly discussed the micro-impact of anti-corruption on enterprises from the perspective of non-productive behavior and productive behavior of enterprises. As most of the literatures suggested, anti-corruption increases the likelihood that companies will be found and punished for building political relationships through non-productive behavior, which will significantly reduce the intrinsic incentive for enterprises to seek political connections (Fan et al. 2007;Cai et al. 2011). Wang and Gao (2017) used on the data of A-share listed companies in China from 2010 to 2013 to establish a fixed effect model and prove that anticorruption will increase the cost of non-productive behavior of enterprises and motivate managers to allocate superior resources of enterprises to productive fields.
For the productive activities of enterprises, some scholars believed that anti-corruption can create a fair and sound system environment, which is conducive to guarantee the expected benefits of enterprise innovation, stimulate enterprises to invest more funds in the field of production, and form innovative development strategy (Berkowitz et al. 2015;Dang et al. 2015;Xu and Yano 2017). Kong et al. (2020) built DID model against the background of China's strong anticorruption campaign since the 18th National Congress of the CPC, and proved that anti-corruption can positively influence enterprise total factor productivity by improving enterprise investment efficiency and increasing enterprise innovation input. This conclusion is consistent with that of Xu and Yano (2017) and Dang et al. (2015).

Comments on literatures
Existing studies contribute to the in-depth understanding of the relationship between anti-corruption and environmental information disclosure, and provide valuable clues for the mechanism and empirical model setting of this paper. However, the following aspects are still worthy of further discussion: Firstly, although some literatures attempt to discuss the influence of macro-institutional environment on environmental information disclosure from the perspective of political relevance, they failed to deeply analyze the internal mechanism of the impact of anti-corruption on environmental information disclosure. This paper carries out theoretical analysis and quantitative research from the perspective of anti-corruption, which is helpful to provide evidence that affects the quality of corporate environmental information disclosure at the macro level.
Secondly, the existing research has not formed a scientific system for the measurement of anti-corruption, and all the important variables affecting anti-corruption and environmental information disclosure cannot be controlled in the model, which are is easy to cause endogenous problems. Therefore, this paper takes the anti-corruption campaign of the 18th CPC National Congress as an exogenous impact to build a DID model, and carries out a series of robust tests, such as parallel trend test and placebo test.

Difference-in-differences model
The basic idea of the difference-in-differences model is to evaluate the effect of the policy by comparing the changes before and after the implementation of the policy between the treatment group and the control group. The specific model can be constructed as follows: In this form, the samples can be divided into four groups: the pre-reform reference group (treat it =0, post it =0), the prereform treatment group (treat it =1, post it =0), the post-reform reference group (treat it =0, post it =1), and the post-reform treatment group (treat it =1, post it =1).
For reference group (treat it =0), the difference between before and after the implementation of the policy can be expressed as β 2 . The specific expression is as follows: For treatment group (treat it =1), the difference between before and after the implementation of the policy can be expressed as β 2 +β 3 . The specific expression is as follows: Therefore, the net effect of the policy can be estimated as β 2 +β 3 -β 2 =β 3 , that is, the coefficient of post it ×treat it . The above model not only control the unobservable individual heterogeneity between samples, but also control the influence of the unobservable overall factors that change over time, and finally obtain an unbiased estimate of the policy effect (Chen et al., 2012).
Significantly, model (1) is used on the premise that all the individuals in the treatment group began to be impacted by the policy at exactly the same time. However, the time of carrying out the anti-corruption campaign in each province is different; model (1) cannot be used to estimate the effect of anti-corruption campaign. To settle this problem, this study draws on the idea of Dang et al. (2015), setting the heterogeneous timing DID model. The heterogeneous timing DID model's essence is to place the different time of anticorruption shocks in each province in the same model, so it can automatically generate the treatment group and the control group. Then, compare the differences of the quality of enterprise environmental disclosure between the treatment and control groups before and after the implementation of the policy to obtain the net effect of the anti-corruption. The specific model is constructed as follows: where i, j, and t represent corporates, cities, and years, respectively. EIDI ijt indicates the corporate environmental information disclosure index of corporate j located in the city i by year t. AntiCorr ijt is a dummy variable for provinces, if a senior government official belongs to province i is investigated in year t, the value of AntiCorr ijt equals one. Otherwise, otherwise it is zero. Control ijt is a vector of control variables used to control the effect of observable factors on corporate environmental information disclosure and the anti-corruption campaign. μ i , γ t , η r , and λ s indicate firm fixed effects, time fixed effects, region fixed effects, and sector fixed effects, respectively. Trend t indicates the time span considered in the equation. ε ijt is an error term. In the above equation, β 1 is the anti-corruption campaign effect that we estimate. In order to alleviate the bias of regression results caused by endogenous problems, the explanatory variables we used in empirical regression all lag one stage in the time dimension.

Sample and data
Considering that the relevant regulations issued in China only makes clear the legal responsibility to require the heavy polluting enterprises to disclose environmental information, while non-heavy polluting enterprises can voluntarily disclose environmental information, we believe that anti-corruption should have no impact on non-heavy polluting enterprises. Therefore, to evaluate the impact of anti-corruption campaign on the quality of enterprise environmental disclosure, we select heavy polluting listed companies in Shanghai and the Shenzhen A-share manufacturing industry over the 2008-2018. Specifically, the identification of pollution-intensive industry is based on the Guidelines for Environmental Information Disclosure of Listed Companies, which was released by the Ministry of Environmental Protection in September 2010, determining steel, cement, coal, petrochemical, thermal power, electrolytic aluminum, metallurgy, building materials, chemical, brewing, papermaking, mining, pharmaceutical, leather, fermentation, and textile industries such as16 kinds of industries for the pollution-intensive industry.
We collect the enterprise Environmental Information Disclosure Index by looking for the Enterprise Annual Report. Corporate financial data such as enterprise-scale and asset-liability ratio are from China's Stock Market Accounting Research Database (CSMAR). Annual GDP per capital and foreign investment were collected from China's Economy Prediction System Database (EPS), and the marketization index compiled by Wang et al. (2017b). After obtaining the original data, we processed it as follows: first, we match enterprise-level data with municipal-level data according to the enterprise's geographic location. Second, in order to avoid the influence of outliers and extreme values of data, we remove ST and *ST companies, and perform 1% winsorize processing on the major continuous variables. Finally, a total of 6364sample data of 914enterprises were obtained.

Variable measurement
Dependent variable: environmental information disclosure Following prior literature (Al-Tuwaijri et al. 2004;Clarkson et al. 2008), we adopt the "content analysis method" to estimate the quality of enterprise environmental disclosure. Specifically, based on the Measures for Environmental Information Disclosure of Enterprises and Institutions and the Guidelines for Environmental Information Disclosure of Listed Companies, combined with the production process of listed manufacturing companies, this paper determines three aspects of "prior prevention inputs," "in-process production control," and "post-event output governance," and sets up 15 secondary projects according to the sequence of occurrence as the evaluation indicators of environmental information disclosure index (Table 1).
In order to improve the scientificity of measurement results, we refer to and modify the evaluation methods of Wiseman (1982) and Darrell and Schwartz (1997), select two quality dimensions of quantification and significance, and the time dimension is removed. This is because the existing literature believes that the disclosure of past, present, and future environmental information of enterprises can comprehensively increase the quality of enterprise environmental disclosure. But in fact, enterprises prefer to disclose only self-interest environmental information. It is difficult for us to verify the future information, so adding time dimension indicators into the index system will often overestimate the scoring results.
In the process of scoring, we adopt the method of twoperson independent scoring. Only when the consistency reaches 90% in the trial stage can the formal scoring begin. If there is a big difference between the two raters in the formal scoring, it will be coordinated by a third person. Finally, we conducted a credibility test on all scoring results, and the value of Cronbach's α was above 0.9, indicating that the scoring results were more reliable.

Independent variable: anti-corruption
Referring to the practice of existing literatures to measure anticorruption (Pan and Tian 2017; Kong et al. 2020), we collect the time of senior government officials who were investigated for corruption in 31 provinces from 2012 to 2015. Table 2 presents the distribution and the year of investigated provinces. If a senior government official in a province is investigated by the Central Inspection Team (CIT) after the 18th CPC National Congress, we believe that the anti-corruption intensity of the province has changed significantly, and then the anti-corruption intensity of the province equals one; otherwise, it is zero, thus constructing a dummy variable of anti-corruption.

Control variable
Control variables are used to mitigate the influence of missing variables on the regression results. In particular, this study selects the enterprise characteristics such as enterprises' size, asset-liability ratio, book-market ratio, etc. as control variables. Besides, annual GDP per capital, marketization index, and foreign investment will be controlled to measure the local economic performance, the intensity of marketization, and the degree of opening to the outside world respectively, which not only affect the anti-corruption efforts, and will differentiate the quality of enterprise environmental disclosure by influencing the concept of corporate environmental governance  The variables descriptive statistics are reported in Table 3.

Results and discussions
Anti-corruption campaign and the quality of the quality of environmental disclosure The estimated results of the anti-corruption campaign on the quality of enterprise environmental disclosure are reported in Table 4. Column 1 of Table 4 shows the impact of AntiCorr ijt on the quality of enterprise environmental disclosure on the basis of controlling time and firm fixed effects. We also control for industry-by-year fixed effects and city-by-year fixed effects to remove any time-variant shocks at the industry and regional level. The regression coefficient of AntiCorr ijt is positively significant at the 5% level. Further, we control the municipal-level characteristics and enterprise-level characteristics in columns 2-3; the impact coefficient of AntiCorr ijt is significantly reduced, which shows that the endogenous problem caused by missing variables has been alleviated to a certain extent. The third column contains all the control variables and fixed effects, and thus, the estimation results are more reliable. It can be found that the coefficient of AntiCorr ijt is significantly positive in the level p < 0.05, and the anti-corruption campaign leads to an average increase of 0.49 points in the environmental information disclosure index, accounting for 5.3% of its average level. This conclusion is consistent with Cole (2007). They both believe that corruption not only directly damages the ecological environment, but also indirectly aggravates pollution by reducing the threshold of environmental access, weakening the intensity of environmental regulation and expanding the scale of informal economy. However, since the 18th CPC National Congress, China has carried out the most powerful anticorruption actions, including the introduction of "eight policies" and "six prohibitions", further strengthening the central inspection work, opening up the Internet reporting channels, etc. The high-intensity anti-corruption action has achieved remarkable results, and a large number of senior government officials and enterprise executives have fallen, which has a deterrent impact on other government officials in office, thus effectively curbing the abuse of power, rent-seeking, and other corruption, and greatly improving the competition environment. In this case, enterprises not only face a greater risk of punishment, but also have to pay higher rent-seeking costs to establish a close relationship with the government. The rational approach of enterprises is to reduce rent-seeking, and take more innovative behavior such as technology improvement and management optimization as important means of competition, so as to improve the quality of enterprise environmental disclosure.

Parallel trend assumption and dynamic effect analysis
Using the double-difference method needs to satisfy the parallel trend assumption; that is, before implementing the anticorruption campaign in China, the changing trend of the quality of enterprise environmental disclosure of the treatment and control group is consistent. According to the framework of Li et al. (2015) and Zhang et al. (2019), this study uses the event study analysis to test the parallel trend before the impact of the anti-corruption campaign and further observes the dynamic effect of the anti-corruption campaign on the quality of enterprise environmental disclosure. The specific model is constructed as follows: Anticorr i,t is a dummy variable in the model (2), representing a senior government official in province i which is investigated in year t. Anticorr i,t-k represents the preposition of period k, which is set to examine whether the enterprises in the treatment group and control group experienced differential trends in enterprise environmental information disclosure before the initial anti-corruption campaign. What we focused on is the coefficient F k of the dummy variable Anticorr i,t-k . The dummy variable can estimate the differences in corporate environmental information disclosure between treatment group and control groups before the initial anti-corruption campaign.
If the corresponding F k is not significant, it means that there is no systematic difference, and the parallel trend hypothesis can be established. The coefficient L j of dummy variable Anticorr i,t+j allows us to assess the dynamic effect of anticorruption on the quality of enterprise environmental disclosure.
The estimated results of the parallel trend assumption are reported in Figure 1. The insignificant coefficients F k indicate that treated and control group have a parallel trend before the initial anti-corruption campaign, which shows the parallel trend hypothesis can be verified. Further observing the dummies variables after the anti-corruption policy, the coefficient L j decreases year by year from 2.99 (L0) to 1.31 (L5), and gradually close to significant in level p < 0.1. This shows that anti-corruption can improve the quality of enterprise environmental disclosure, but the effect is difficult to sustain. The reason may be that the Party Central Committee of  China has a "zero tolerance" attitude towards corruption, which has a deterrent effect on government officials and greatly reduces the non-productive behavior of enterprises such as rent-seeking. Therefore, the phenomenon of corruption will gradually disappear, the intensity of anti-corruption will also decrease, and the effect on the quality of enterprise environmental information disclosure will gradually weaken.

Robustness check
First, replace the interpreted variable. This paper selects the number of anti-corruption cases investigated and prosecuted in each province every year to replace the existing explanatory variables for robustness test (Dang et al. 2015). The regression results are shown in Table 6 column 1. The number of corruption and malfeasance cases investigated and prosecuted has a positive impact on the quality of enterprise environmental information disclosure at the level of 1%. Secondly, replace the clustering robust standard error. Because the more robust standard error can reduce the deviation of statistical inference, this paper, referring to the existing literature practice, replaces the enterprise level clustering robust standard error with "city-enterprise" higher dimensional clustering robust standard error. The regression results are shown in Table 6 column 2. Anti-corruption has a positive impact on the quality of enterprise environmental information disclosure at the level of 10%. It can be seen that setting different clustering robust standard errors will not affect the conclusion of this paper.
Finally, use PSM -DID model. We have added controlling variables in the model (2) to alleviate the self-selection problem.. However, OLS requires a priori to give the conditional expectation function form of dependent variables. in the treatment group and control groupWhenthe conditional expectation function does not conform to the reality or the economic theory, the extrapolation bias caused by the model setting error will be generated, and the average causal effect cannot be obtained. Therefore, we introduce propensity score matching (PSM) to test further the robustness of the causal relationship between the anti-corruption campaign and the quality of enterprise environmental disclosure.
In particular, the controlling variables in benchmark regression are selected as covariates and EIDI is taken as the dependent variable. The propensity score is calculated through the Logit model. According to the propensity score, the treated and control group are matched, so as to calculate the average treatment effect for the treated. In this paper, the radius matching method is used to verify the robustness of the results, and the absolute value of the standardized deviation of most covariates is less than 10%, which is significantly lower than that before matching, and shows that the matching effect is well (Table 5).
After completing the above matching, the model (1) is used for regression again to identify the net effect of the anticorruption campaign on the quality of enterprise environmental disclosure. The estimation results of PSM-DID are reported in Table 6 column 3. After controlling the observable systematic deviation, the anti-corruption campaign has a significant positive impact on the quality of enterprise environmental disclosure in the level p < 0.05, which is consistent with the baseline regression results.

Placebo test
Although the regression results above control the systematic differences caused by observable variables such as the total number of employees and asset-liability ratio, we are still Note: * , ** , and *** indicate that the estimated results are significant at the levels of 1%, 5%, and 10%, respectively, respectively; the clustering robust standard deviations are in parentheses Fig. 1 The parallel trend of EIDI. Note: This figure plots the period-byperiod estimated coefficients obtained through model (3), and the vertical line indicates the 90% confidence interval of the estimated coefficient concerned that the unobservable systemic factors may interfere with the regression results. Therefore, our research intends to construct a series of counterfactual frameworks for placebo test. If the impact of anti-corruption campaign is still positively significant under the counterfactual circumstances, it means that this utility comes from the unobservable systemic factors, not from the implementation of anti-corruption campaign.
We first use bootstrap to randomly allocate the anti-corruption campaign's selection time for each city, and repeat the regression 500 times according to model (1). The estimated results are reported in Figure 2, which can be found that the t-value of the coefficient is approximately normally distributed, mostly around 0, rarely around ±3 and ±4. It means that in the 500 random experiments conducted, the hypothetical policy implementation will have a small probability that the regression coefficient of EIDI will be significantly positive and negative. Therefore, the counterfactual treatment effect of the implementation of the anticorruption campaign does not exist.
What's more, we further assumed that advancing anticorruption campaign policy 1 year to 6 years, respectively, to construct the pseudo-implementing time of the anti-corruption campaign policy. If the coefficient is not significant, the anticorruption campaign policy indeed increases the corporate environmental information disclosure. Otherwise, basic regression is not robust. The bootstrap results are reported in Table 7. The estimated results indicate that the anti-corruption campaign policy implemented 1 year in advance positively affects the environmental information disclosure in level p < 0.05. The possible reason is that as early as the early stage of reform and opening up, China has been clear about the road of anti-corruption within the party; the work plan for establishing a sound system for punishing and preventing corruption issued by the Party Central Committee in 2008 makes specific plans for anticorruption actions in the future. The coefficient of the anticorruption campaign policy 2 to 6 years in advance does not significantly influence the quality of enterprise environmental disclosure, which shows the robustness of the basic regression.

Influential mechanism
Through literature review, we have reason to believe that there is also a causal relationship between anti-corruption and environmental information disclosure. In particular, by strengthening the punishment of government officials' illegal behavior and strictly supervising the government's administrative power, anticorruption has greatly increased the cost of non-productive behavior of enterprises. More importantly, anti-corruption clearly standardizes administrative rules and procedures, limits the government's discretionary power, and reduces the government's excessive intervention in the market, which helps create a fair institutional environment. At this time, enterprises are more willing to invest capital in productive behaviors such as innovation, so as to promote green economic development and improve the quality of environmental information disclosure. To sum up, anti-corruption will reduce enterprises' non-productive behaviors, encourage enterprises to make productive investments, and ultimately improve the quality of enterprise environmental information disclosure.
Based on the above theoretical analysis, this study selects two mediating variables of innovation input and violation cost, and uses the stepwise method to test the mediating effect, so as to test whether the anti-corruption campaign will influence the non-productive and productive behaviors of enterprises, and further differentiate the disclosure of enterprise environmental information. The specific model is constructed as follows: In Equations (4) and (5), mediation it is the mediating variable, including innovation input and violation cost. Among them, innovation input is measured by total R&D investment, which can timely reflect the enterprise innovation investment in the short term; violation cost is measured by entertainment and travel expenses. When the cost of violation increases, the cost of rent-seeking should be reduced.
The specific test steps of mediating effect are as follows: step 1 is to confirm that anti-corruption campaign will indeed affect the quality of enterprise environmental disclosure. That is, based on the model (2) coefficient, β 1 is significant, and the coefficient θ med in Equation (4) and the coefficient φ med in Equation (5) are tested in turn. If both are significant, it means that the indirect effect exists and is tested in step 3. If at least one is not significant, the second step test is conducted.
Step 2 uses the Sobel method to test the original hypothesis directly: θ med × φ med =0, if significant; the mediating effect is significant, step 3; otherwise, stop the analysis.
Step 3 is to test the coefficient φ CR in Equation (5); if not significant, then the direct effect is not significant, indicating that the model only exists mediating effect, and if significant, then step 4 test is needed.
Step 4 compares the symbols of θ med × φ med and φ CR , if the symbol is consistent, it means that there is a partial mediation effect, and the test results are as follows.

Violation cost effect
Violation cost mechanism test results are reported in Table 8.
Step 1 (Table 8, column 1) indicates that the dummy variable AntiCorr ijt has a negative impact on the entertainment and travel expenses, but it is not statistically significant. The results of Table 8, column 2, show that entertainment and travel expenses have a negative effect on the quality of enterprise environmental disclosure at the 10% level. Therefore, this study adopts the Sobel method to test Step 2. The results show that the Z statistic value of EIDI is −2.911, which rejects the original hypothesis at the level of 5%, confirming the establishment of violation cost mechanism. The explanation for this is that the strengthening of anti-corruption makes the government's power and behavior subject to strict regulations, which greatly increases the probability of non-productive behavior of enterprises being found and punished. Enterprises need to pay higher rent-seeking costs to maintain a stable relationship with the government in order to avoid environmental regulation. At this time, with the reduction of rentseeking benefits and the increase of rent-seeking costs, the possibility of enterprises engaging in rent-seeking is reduced, and enterprises get internal incentives to improve the quality of environmental information disclosure.
The results of step 3 show that (Table 8, column 2) the dummy variable AntiCorr ijt has a positive influence on the quality of enterprise environmental disclosure in the level p < 0.1, and is consistent with the symbol of θ med × φ med , indicating that violation cost has a partial mediating effect. Note: * , ** , and *** indicate that the estimated results are significant at the levels of 1%, 5%, and 10%, respectively; the clustering robust standard deviations are in parentheses Fig. 2 Bootstrap result

Innovation input effect
Innovation input effect mechanism test results are reported in Table 8.
Step 1 (Table 8, column 3) indicates that the dummy variable AntiCorr ijt has a positive impact on the total R&D investment, but it is not statistically significant. The results of Table 8, column 4, show that corporate total R&D investment has a positive impact on the quality of enterprise environmental disclosure in the level p < 0.1. Therefore, this paper adopts Sobel method to test step 2. The results show that the Z statistic value of EIDI is 3.178, which rejects the original hypothesis at the level of 5%, confirming the establishment of a technological innovation mechanism. This conclusion is consistent with Xu and Yano (2017), which means that the anticorruption campaign has created a fair and sound institutional environment, stimulated enterprises to invest more funds in the production field, developed innovative activities such as technology research and development, and improved enterprises' technological level. The improvement of enterprise technology level can optimize enterprise production structure, improve energy efficiency, save energy consumption, reduce pollutant emissions, and improve enterprise environmental performance. Based on this, enterprises will be more ready to disclose high-quality environmental information to respond the supervision of stakeholders.
The estimated results of step 3 show that (Table 8, column 4) the dummy variable AntiCorr ijt has a positive effect on the quality of enterprise environmental disclosure in the level p < 0.01, and is consistent with the symbol of θ med × φ med , indicating that management optimization has a partial mediating effect.

Heterogeneity tests
The above theoretical analysis and empirical results confirm that the anti-corruption campaign can encourage the improvement of the quality of enterprise environmental disclosure. However, the above analysis is based on the fact that the anti-corruption campaign has the same impact on all listed manufacturing enterprises, ignoring institutional performance differences. In fact, the effect of anti-corruption on the quality of enterprise environmental disclosure is often different for enterprises which have different property rights, enterprise size, and regions. Specifically, enterprises with different property rights and sizes often have differences in the degree of association between resources and government, and taking into account the different levels of corruption in different regions, making different types of enterprises subject to different constraints of anti-corruption, thereby affecting the quality of enterprise environmental disclosure. Therefore, we further Note:(1) * , ** , and *** indicate that the estimated results are significant at the levels of 1%, 5%, and 10%; (2) control variables include enterprise-level and municipal-level; (3) fixed effects include year fixed effect, firm fixed effect, city-by-year fixed effect, and industry-by-year fixed effect. (4) PRE_1-PRE_6 represents the policy which is 1-6 years ahead of schedule, respectively discuss the relationship between the anti-corruption campaign and the quality of enterprise environmental disclosure from three aspects: enterprise nature, enterprise size, and environmental supervision strength

Enterprise nature
This study divides enterprises into state-owned enterprises and non-state-owned enterprises according to different property rights and further verifies the impact of the anti-corruption campaign on enterprise environmental information disclosure under different property rights. The regression results are reported in column 1-2 of Table 9. The estimated results indicate that the dummy variable AntiCorr ijt has a positive influence on the quality of enterprise environmental disclosure of non-state-owned enterprises in the level p < 0.1, and the regression coefficient is 0.505, but it has no statistical significance for the environmental information disclosure quality of state-owned enterprises. This means that the anti-corruption campaign has a greater role in promoting the quality of enterprise environmental disclosure of non-state-owned enterprises.
The estimated results indicate that the anti-corruption campaign has created a fair institutional environment for stateowned enterprises and non-state-owned enterprises, and government power has been effectively supervised and restricted. In this institutional environment, non-state-owned enterprises can rely on the price competition mechanism to obtain more resources in the market, which can further encourage nonstate-owned enterprises to carry out technological innovation and management optimization, so as to improve the corporate environmental performance and increase the quality of enterprise environmental disclosure. In addition, non-state-owned enterprises often take profit maximization as their business philosophy and face greater survival pressure. They are looking forward to obtaining competitive advantage by improving the quality of enterprise environmental disclosure, and thus the innovative activities of non-state-owned enterprises are more adventurous. However, the state-owned enterprises draw more attention to stability and efficiency in strategy and tend to adopt a defensive attitude. Their diversified development goals make the technological innovation lack internal motivation, the innovation incentive effect of anticorruption is smaller so that the quality of environmental information disclosure is difficult to improve (Cull and Xu 2005). Therefore, the anti-corruption campaign has a greater role in increasing the quality of environmental information disclosure of non-state-owned enterprises.

Enterprise size
Referring to most literature practices, this paper considers total assets as the classification basis and defines small-and medium-sized enterprises as those less than the median of total assets, and large enterprises as those greater than the median. The estimated results are reported in column 3-4 of Table 9. The estimated results indicate that the interaction item positively influences the environmental information disclosure of enterprises of different sizes in the level p < 0.1, which is consistent with the benchmark regression results. Further observation of the regression coefficient of each group indicates that the regression coefficient of the quality of large enterprise environmental disclosure is greater than that of small-and medium-sized enterprises, which indicates that the anticorruption campaign will improve the quality of enterprise Note: * , ** , and *** indicate that the estimated results are significant at the levels of 1%, 5%, and 10%, respectively; the clustering robust standard deviations are in parentheses environmental disclosure of different sizes, and the positive effect of the anti-corruption campaign on the environmental information disclosure quality of large enterprises is greater. This paper argues that small-and medium-sized enterprises are less associated with the government and have fewer production resources to obtain. Anti-corruption is conducive to breaking this unfair competitive environment and ensuring the input and income of small-and medium-sized enterprises in production and management activities to a greater extent. Their willingness to carry out technological innovation and management optimization will significantly increase, and the quality of enterprise environmental disclosure will also be improved. For large enterprises, anti-corruption makes it difficult to obtain additional production resources through non-productive activities such as rentseeking, which greatly weakens large enterprises' competitive advantage. Therefore, it will force large enterprises to innovate and continuously improve the quality of enterprise environmental disclosure to obtain environmental premium to maintain their competitiveness. However, it is worth noting that compared with small-and medium-sized enterprises, large enterprises have the advantages of advanced technology and high-end technical talents, which can innovate at a low cost. Therefore, anti-corruption has a stronger promoting effect on large enterprises to improve the quality of enterprise environmental disclosure.

Environmental supervision strength
Referring to the existing literature practice, this research uses the number of environmental administrative punishment cases of each province in 2011 to measure the local environmental supervision strength (Ren et al. 2020). We define the provinces with the number of environmental administrative punishment cases higher than the median as the provinces with higher environmental supervision strength, and vice versa as the provinces with lower environmental supervision strength. The regression results are reported in column 5-6 of Table 9.
The estimated results indicate that the interaction item has a positive impact on the environmental information disclosure of enterprises in the areas with strong supervision in the level p < 0.05, and the regression coefficient is 1.353, but it has no statistical significance for the quality of enterprise environmental disclosure of enterprises in the areas with weak supervision. The results show that anti-corruption plays a more significant role in promoting enterprise environmental information disclosure in weak environmental supervision areas.
The explanation for this is that the corruption problem may be more serious in weak environmental supervision areas. It is easy for environmental protection departments to collude with local-related polluting enterprises and take strategic interaction. The environmental regulations of the government are completely invalid, and it is difficult for enterprises to obtain incentives to disclose high-quality environmental information (Sheng et al. 2019). At this time, the enhancement of anticorruption makes it difficult for enterprises to break through the environmental information regulation through rentseeking and collusion, forcing enterprises to improve technology and management and ultimately improve the quality of enterprise environmental disclosure. However, the areas with stronger environmental supervision have a lower tolerance to corruption, and anti-corruption has no significant incentive effect on innovation, so it is difficult to influence the quality of enterprise environmental disclosure. Therefore, this paper argues that in areas with weak environmental supervision, anti-corruption plays a greater role in improving corporate the quality of enterprise environmental disclosure.

Conclusions and policy suggestions
Based on the quasi-natural experiment of the anti-corruption campaign, this research uses the data of environmental information disclosure index of heavy polluting enterprises in Note: * , ** , and *** indicate that the estimated results are significant at the levels of 1%, 5%, and 10%, respectively; the clustering robust standard deviations are in parentheses Shanghai and Shenzhen A-share manufacturing industry over 2008-2018 to construct a double difference model (DID) to estimate the net effect of the anti-corruption campaign on the quality of enterprise environmental disclosure. The main conclusions are as follows: firstly, the DID regression results indicate that the anti-corruption campaign has a significant positive influence on the quality of enterprise environmental disclosure. The regression results are tested by a series of robustness tests such as parallel trend test and placebo test. Secondly, this paper opens the black box of the relationship between the anti-corruption campaign and the quality of enterprise environmental disclosure through mechanism test and confirms that anti-corruption will increase the violation costs, encourage enterprises to carry out innovation activities, and ultimately improve the quality of enterprise environmental disclosure. Finally, the heterogeneity analysis results show that non-state-owned enterprises, large enterprises, and regions with stronger environmental supervision will improve the quality of enterprise environmental disclosure to a greater extent. Based on the analysis, we offer the following policy recommendations: Firstly, the incentive function of anti-corruption on the quality of environmental information disclosure, especially in the regions with stronger environmental supervision, further confirms the excellent achievements of the anticorruption action of the CPC Central Committee from the micro perspective. In order to further consolidate the effect of anti-corruption, government departments should constantly strengthen the restrictions on the power of government officials, and severely punish government officials who violate the rules of corruption, such as overstepping their powers and abusing their powers.
Secondly, considering that the impact of anti-corruption on non-state-owned enterprises and small-and medium-sized enterprises' environmental information disclosure is more significant, this paper holds that the degree of government intervention in resource allocation should be weakened, so that non-state-owned enterprises and small-and medium-sized enterprises can grow in a fairer competitive environment.