3.1 Study design

This study takes the disclosure policy of PITI in 113 cities in China jointly published by the IPE and the NRDC as a quasi-natural experiment, and uses DID methods to identify the causal effect between public supervision and disclosure of enterprises' environmental violations. That is to say, in this paper, the cities that implement the disclosure policy of PITI were set as the treatment group, whereas the cities that do not implement the pilot served as the control group. The specific DID model is:

(1)\begin{equation}\begin{aligned} Corporate\;violatio{n_{it}} & ={\alpha _0} + \beta PIT{I_i} \times Tim{e_t} + \sum Contro{l_{it}}\\ & \quad + {\mu _i} + {\delta _t} + {\varepsilon _{it}}, \end{aligned}\end{equation}

where Corporate violation_it represents the level of enterprises' environmental violation of city i in the t year, measured by regional total number of enterprises with environmental violations and percentage of total number of enterprises with environmental violations in total regional enterprises, respectively. PITI_i indicates whether city i implements the disclosure policy of PITI. If yes, the value is 1, otherwise, it is 0. The variable Time_t indicates whether the time is after 2008 or not, and if it is after 2008, the value is 1; otherwise, it is 0. δ_t is the time fixed effects, μ_i is the individual fixed effects in the city, ε_i is the error term, and Control_it is the selected series of control variable. The coefficient β is the focus of research. If β > 0, it means that public supervision policy implementation is conducive to enterprises' environmental violation disclosure.

However, due to obvious differences between cities' environmental standards, and government-enterprise relations, these factors may affect the role of public supervision in disclosing enterprises' environmental violations, resulting in heterogeneity of results. In order to carry out the necessary empirical test on this analysis conclusion, we establish the following model:

(2)\begin{equation}\begin{aligned} Corporate\; violation{\; _{it}} & ={\alpha _0} + \gamma PIT{I_i} \times Tim{e_t} \times leve{l_i} + \beta PIT{I_i}\\ & \quad\times Tim{e_t} + \sum Contro{l_{it}} + {\mu _i} + {\delta _t} + {\varepsilon _{it}} \end{aligned}\end{equation}

where level_i indicates the difference in the environmental standards, and government-enterprise relations of city i. In the subsequent empirical tests, we distinguish the two with different standards to examine the heterogeneity of the role of public supervision in revealing enterprises' environmental violation in terms of environmental standards and government-enterprise relations, respectively. The coefficient γ is used to evaluate the heterogeneity effect of public supervision in disclosing enterprises' environmental violation under different factors.

Furthermore, the parallel trend assumption must be met. It assumes that the number of enterprises' environmental violations between the treatment group and the control group will not change significantly over time without the implementation of environmental information publicity. If there is a significant difference, the parallel trend hypothesis will not be met, and the regression results will be biased. Based on the above reasons, this paper expands on the basis of model (1) and establishes a dynamic effect model, as follows:

(3)\begin{equation}\begin{aligned} Corporate\; violatio{n_{it}} & ={\alpha _0} + \sum\nolimits_{\tau \in \{{2004,2005, \cdots,2010,2011} \}} {{\theta _\tau }PIT{I_i} \times yea{r_\tau }} \\ & \quad + \sum Contro{l_{it}} + {\mu _i} + {\delta _t} + {\varepsilon _{it}}. \end{aligned}\end{equation}

Parameter ${\theta _\tau }$ represents a set of evaluation effects, which indicates the role of public supervision of city i in disclosing environmental pollution violations in year τ, year_τ represents the dummy variable for each year, so as to examine dynamic effects of public supervision on enterprises' environmental violation disclosure. If the coefficients ${\theta _{2004}},\; \; {\theta _{2005}},\; \; {\theta _{2006}}\; \textrm{and}\; {\theta _{2007}}$ are not significant, it means that there is no systematic difference between the treatment group and the control group before policy implementation; otherwise, it means that there is a systematic difference, that is, the parallel trend assumption required by the DID is not met.

3.2 Variables

3.2.3 Control variables

At the macro regional level, enterprises' environmental violations are not only driven by factors such as economic development, but are also restrained by regional marketization, public education level, and openness to the outside world. On the one hand, local governments sacrifice the environment and allow enterprises to engage in environmental pollution to actively develop the economy. On the other hand, the pressure of the public opinion and the improvement of environmental awareness may force the government to curb enterprises' environmental violations. Considering the above two factors, the GDP per capita, urbanization level, financial pressure and tax burden intensity are selected as control variables to measure the level of economic development, while the degree of openness to the outside world,Footnote ⁵ education level, market structure, and financial development are used to measure the degree to which an area attaches importance to environmental quality.

At the micro enterprise level, existing studies have shown that factors such as asset-liability ratio, scale, and owner's equity influence enterprise output, which is why the total assets are used in this study to measure an enterprise's scale, the logarithm of total enterprise debt to measure debt, the logarithm of total industrial sales value to measure output, time of enterprise establishment to measure age, and the ratio of the owner's equity to total industrial output value to measure the owner's equity of the enterprise. Additionally, the logarithm of administrative expenses is used to represent an enterprise's cost, the logarithm of the enterprise's payable income tax to illustrate the enterprise's tax burden, and the percentage of export delivery value to the total enterprise's industrial output value to illustrate the enterprise's export capacity. At the same time, the impact of the subsidy income on the enterprise's output is also taken into account. The variable selection is detailed in table 1.

Table 1. A description of specific variables and the descriptive statistics

Notes: The Environmental Violation indicators of enterprises at the urban level are selected from 2004 to 2012, the economic indicators at the urban level are selected from 2004 to 2012, which are derived from the China Urban Statistical Yearbook. The indicators at the enterprise level are selected from 1998 to 2011, which are derived from the China industrial enterprise database.

3.3 Data

Taking the disclosure policy of PITI as a quasi-natural experiment, this study investigates the effect of public supervision on enterprises' environmental violation disclosure. To address the fact that the literature has only focused on analyzing the environmental benefits at the macro level, we combine data at the regional level with data on enterprises at the micro level to detail the path of achieving environmental benefits through public supervision. The specific data are as follows.

First, this study starts at the macro regional level, taking 286 prefecture-level cities in China as the research object and using urban panel data from 2004 to 2012. The original data are from the statistical yearbooks of Chinese cities. The investigation period was selected from 2004 to 2012 because the data of enterprises' environmental violations can only be obtained from 2004 to 2012, which limits the sample period. In particular, the data on environmental violations mainly come from public enterprise environment reports published by the Research Center for Public and Environmental Affairs, a non-profit environmental organization, which released information on those enterprises whose water pollution exceeds the standard level. We manually collated the number of enterprises in each city that were disclosed as having excessive water pollution emissions between 2004 and 2012. The specific data descriptions are shown in figure 2.

Figure 2. The time and regional distribution of enterprises with environmental violation behaviors.

Second, to investigate the impact of public supervision on enterprises' environmental violation disclosure, we used the disclosure policy of PITI to construct a quasi-natural experiment. Thus, the cities that implement the disclosure policy of PITI were set as the treatment group, whereas the cities that do not implement the pilot served as the control group.Footnote ⁶ Considering that the evaluation results are potentially affected by endogeneity, which can lead to biased estimation, we manually collected the number of public petitions on urban environmental issues from the China Environmental Statistics Yearbook.Footnote ⁷ And this number of public petitions was set as the instrumental variable. To eliminate the interference of other policies over the analysis period, we searched the websites of governmental agencies for the carbon pilot policy of ‘five provinces and eight cities, the “regional limited approval”, and “watershed limited approval” implemented by the State Environmental Protection Administration in some regions,Footnote ⁸ as well as the carbon trading pilot policy implemented in Beijing, Shanghai, Chongqing, Tianjin, Shenzhen, Guangdong, and Hubei’.

4.1 The impact of public supervision on enterprises' environmental violation disclosures

First, model (1) is regressed to investigate the effect of public supervision on enterprises' environmental violation disclosures. The results are shown in table 2. Column (1) shows the effects of public supervision on the regional total number of enterprises that disclose environmental violations, while column (3) shows the effects of public supervision on the relative number of enterprises that disclose environmental violation. Regardless of whether enterprises' environmental violation disclosures are measured by their total or relative numbers, public supervision promotes environmental violation disclosure, and this promotion effect is significant at the 5 per cent confidence level. Specifically, if other factors remain unchanged, the regression coefficients in columns (1) and (3) in table 2 show that, compared with regions without a public supervision system, a public supervision policy increases the regional total number of enterprises' environmental violation disclosure by 33.86 per cent and the relative number by 5.12 per cent.

Table 2. The impact of public supervision on enterprises' environmental violation disclosure

Note: The values in brackets are robust standard errors of clustering at the prefecture level.

Columns (1) and (3) in table 2 show the results for all cities. Considering that cities disclosing information publicity are distributed throughout the eastern, middle, and western regions, factors such as city size may cause large differences between the cities in the treatment and control groups, which may cause biased evaluation results and violate the parallel trend assumption. Therefore, to reduce this evaluation bias, we also use propensity score matching for calibration. The specific matching ideas are as follows. First, taking whether city publicizes environmental information as the dependent variable and using the economic growth rate, labor endowment, government size, and other factors as independent variables, the probit model is used to calculate propensity scores. The probit model is as follows:

\[Pr(\textrm{Contro}{\textrm{l}_i}) = Pr(PITI = 1|\textrm{Contro}{\textrm{l}_i}).\]

$Pr\,\textrm{(Contro}{\textrm{l}_i}\textrm{)}$ represents the prediction probability that the city is selected to implement the environmental information disclosure policy.

Second, according to the calculated propensity scores, the nearest neighbor matching method is used to find cities in the control group that are similar to the sample cities in the treatment group. The specific match is as follows:

\[\begin{aligned} ATT & =E\{{E[{ {Corporate\; violatio{n_{1i}}} |PITI = 1,Pr({Contro{l_i}} )} ]} \\ & \quad\ \ - {E[{ {Corporate\; violatio{n_{0i}}} |PITI = 0,Pr({Contro{l_i}} )} ]} \}. \end{aligned}\]

After matching, the cities that failed to be matched are eliminated and we conduct a DID analysis on the successfully matched sample. The specific results are shown in columns (2) and (4) of table 2. The public supervision system still promotes enterprises' environmental violation disclosure significantly.

4.2 Heterogeneity analysis of impact of public supervision on enterprises' environmental violation disclosure

From table 2, the public supervision system is conducive to revealing enterprises' environmental violations; however, considering the differences in environmental awareness and the government–enterprise relationship between cities, which may affect the role of public supervision, we regress model (2) to study the influence of the above factors, as shown in tables 3 and 4.

Table 3. Impact of public supervision on enterprises' environmental violations under different environmental standards

Note: The values in brackets are robust standard errors of clustering at the prefecture level.

Table 4. Impact of public supervision on enterprises' environmental violations under government-enterprise collusion

Note: The values in brackets are robust standard errors of clustering at the prefecture level.

First, the implementation effect of the public supervision system depends not only on regional legal systems construction, but also on urban environmental standards. That is, in cities with more stringent environmental standards, enterprises may have fewer environmental violations; conversely, looser environmental standards will lead to more obvious environmental violations. In short, these differences likely lead to different effects of public supervision. To test this assumption, this study divides all sample cities into first- and second-class cities according to the city level. Based on China's urban economic development stage and environmental governance, the environmental standards of first-class cities are higher than those of second-class production and so is the degree of environmental standards implementation. The results are shown in table 3. Columns (1)–(2) show the effect of public supervision on the regional total number of enterprises with environmental pollution violations disclosures in the different regions, and columns (3)–(4) investigate the effects of public supervision on the relative number of enterprises with environmental violation disclosure in different regions. Regardless of whether the total number or relative number of enterprises is used to measure environmental violation disclosure, the public supervision in regions with lower environmental standards is conducive to enterprises' environmental pollution violations disclosures, the effect being greater than in regions with higher environmental standards.

Second, the public supervision system is mainly advocated and implemented by the government, while environmental pollution violations are decided by the enterprises. Under the premise of changes in their external environment, enterprises may seek political assistance in the face of pressure due to public supervision. Enterprises form coalitions with local officials through bribery and other means to pursue benefits, while government officials may relax regulations and collude with enterprises for economic benefits. Therefore, government–enterprise collusion may weaken the role of public supervision in revealing enterprises' environmental pollution behaviors. To further test the impact of government–enterprise relationships, we classified the number of environmental pollution violations disclosed by enterprises in different cities and divided them into two categories: ‘best’ and ‘good’. The more enterprises violating environmental pollution regulations are disclosed, the lower the degree of collusion between government and enterprises in the city and the higher the incorruption of government officials. On the contrary, cities where the number of enterprises violating environmental pollution regulations are less disclosed may have a higher degree of collusion between government and enterprises, and incorruption of government officials is low. The results are shown in table 4.

The implementation of a public supervision system in regions with a higher government incorruptibility index has a stronger disclosure effect on enterprises' environmental violations than in regions with a lower government incorruptibility index. This shows that the effect of public supervision on enterprises' environmental violation disclosures increases with the strengthening of government–enterprise relationships. That is, the higher is the incorruptibility index, the lower is the probability of collusion between government and enterprises, the greater is the effect of public supervision in revealing enterprises' pollution behavior; moreover, the smaller is the incorruptibility index, the greater is the probability of collusion between government and enterprise and the lower is the effect of public supervision in revealing enterprises' pollution violations.

5.1 The mechanism of public supervision to expose enterprise environmental violations

The above analysis has shown that public supervision can significantly disclose enterprise environmental violations. Then, what is the internal mechanism of environmental information publicity to improve the public's disclosure of enterprise environmental violations? In order to explain this mechanism, this paper analyzes the contents of China's implementation of environmental information publicity policy, and believes that the internal mechanism of public supervision to expose enterprise environmental violations focuses on the following two aspects.

From a personal perspective, environmental information publicity can make residents more truly understand the current situation of environmental pollution in the city where they live. Serious environmental pollution will make residents aware that their health will be threatened. Therefore, in order to protect their own health, residents will actively supervise the environmental pollution emissions of enterprises, especially the pollution emissions of key enterprises in the city with environmental information publicity. Once the enterprise is found to have environmental pollution violations, the residents will complain to the government department, and finally the enterprise with environmental pollution violations will be publicized in the society.

From a social perspective, during the implementation of the environmental information publicity policy, the objective is not only to disclose the environmental pollution information of the pilot cities, but also to publicize the concept of public participation in environmental governance to actively promote and popularize public participation in urban environmental pollution governance. On the basis of this publicity and promotion, the public will directly participate in urban environmental governance through the supervision of environmental violations of enterprises, timely report environmental pollution violations of enterprises, and reduce urban environmental pollution emissions.

Therefore, whether for personal interests or social welfare, the public will disclose environmental pollution violations of enterprises through supervision, reduce urban environmental pollution emissions and improve environmental quality.

5.2 Unexpected effects of public supervision

The above empirical tests fully demonstrate that public supervision policies can effectively reveal environmental violations. With the gradual implementation of these policies, environmental quality should be greatly improved, which is also the ultimate goal of environmental governance. This section analyzes whether public supervision can help improve regional environmental quality and the internal mechanisms, thus providing important implications for effective public supervision policies.

First, we test the improvement of environmental quality by public supervision to answer the first research question. To measure environmental quality, we refer to the literature. Namely, we select urban industrial waste water discharge (water), industrial smoke emissions (fog), industrial dust emissions (dust), and industrial sulfur dioxide emissions (so2) to measure regional environmental pollution status, and use the DID method to assess the impact of public supervision on environmental quality. The results are shown in table 5. Regardless of which indicators are used to measure regional environmental pollution, the public supervision system has a significant negative effect on environmental pollution emissions. That is, public supervision policies effectively improve environmental quality in a jurisdiction and achieve the expected environment quality.Footnote ⁹ The explanation for this result is that public supervision makes a large number of environmental pollution violation enterprises be disclosed, while in China, the direct result of disclosure is the shutdown and rectification of enterprises, which ultimately reduces the pollution emissions of cities.

Table 5. Impact of public supervision on urban environmental pollution emissions

Note: The values in brackets are robust standard errors of clustering at the prefecture level.

The results in table 5 show that public supervision effectively reduces the different types of pollution emissions and improves regional environmental quality. However, for the internal mechanism for environmental quality improvement, we focus on the path of the public supervision system to achieve environmental benefits. The public supervision system can effectively reveal enterprises' environmental violations, rectifying and reforming these enterprises, which is mainly reflected in an enterprise's output and benefits. To test this, we select the enterprise's output scale (productscale), total income (income), operating income (revenue), and operating profit (profit) to measure the enterprise's output and income, and the results are shown in table 6. Compared with the regions without public supervision policies, enterprises' output and income in policy implementation regions decline significantly.

Table 6. Impact of public supervision on enterprise's output

Notes: (1) The values in brackets are robust standard errors of clustering at the prefecture level; (2) Control variables include size, debt, right, firmtax, age, cost, export, subsidy.

Although table 6 shows the mechanism of public supervision to achieve environmental benefits, considering differences in the pollution discharge of different types of enterprises, we believe that the impact of public supervision on enterprises' output and income can differ. To further test the above mechanism, we divide enterprises according to the degree of environmental pollution into strongly and weakly polluting enterprises, and examine the impact of public supervision on the output and income of these enterprise types. The results are shown in table 7. Columns (1)–(4) examine the impact of public supervision on the output and income of strongly polluting enterprises, while columns (5)–(8) analyze the impact of public supervision on the output and income of weakly polluting enterprises. From the regression results, regardless of whether an enterprise is strongly or weakly polluting, public supervision has a significant inhibiting effect on enterprise output. However, from the coefficients, the inhibiting effect of public supervision on the output of strongly polluting enterprises is higher than for weakly polluting enterprises. This indicates that the reason for environmental quality improvement lies in enterprises' environmental violations disclosures, which can reduce both output and emissions. In addition, this inhibiting effect is more obvious for strongly polluting enterprises.

Table 7. Impact of public supervision on different types enterprise's output

Notes: (1) The values in brackets are robust standard errors of clustering at the prefecture level; (2) Control variables include size, debt, right, firmtax, age, cost, export, subsidy.