Performance Evaluation of the Digital Governance of Water Pollution: A Dual Perspective of Digital Monitoring and Digital Administration

Abstract

As digital technologies advance rapidly, their applications to environmental governance are becoming increasingly widespread. Against this backdrop, we focus on the critical area of water pollution control. Using a fixed-effects model, we empirically examine the impact of digital monitoring and administration on corporate water pollution emissions and the mechanisms involved. The findings are as follows: (1) The digital monitoring of water pollution is significantly positively correlated with corporate water pollution emissions, indicating that while digital monitoring enhances pollution detection capabilities, it has not yet effectively curbed pollution emissions. (2) Digital administration significantly reduces corporate water pollution emissions. (3) Digital administration positively moderates the relationship between corporate technological innovation and water pollution, whereas the moderating effect of digital monitoring is not significant. (4) The heterogeneity analysis reveals that the pollution-reduction effects of digital monitoring and digital administration vary significantly with respect to the corporate ownership structure and industry competitiveness. This study provides theoretical foundations and practical insights for improving digital environmental regulation policies.

1. Introduction

Traditional approaches to water pollution governance have often fallen short of reducing pollution, due to inefficiencies in interdepartmental coordination, inadequacies in institutional design, and challenges in effectively monitoring pollution data [1]. Chapman and Sullivan [2] stressed that the lack of a regular and sustained monitoring, intervention in, and management of water health can result in irreversible harm to human well-being. Developing innovative governance strategies and constructing multi-stakeholder collaboration mechanisms to mitigate water pollution have long been focal points of academic and policy discussions. In recent years, the rapid advancement of digital technologies—particularly big data analytics, IoT-based monitoring, and machine learning—has introduced new tools and methods to enhance the efficiency and quality of water pollution governance [3,4,5,6]. The application of digital technologies to water pollution governance has garnered widespread attention from governments and international organizations [7]. China is no exception. Since 2019, a series of policies, including the “Guiding Opinions on Accelerating the Advancement of Smart Water Resources” and the “Comprehensive Plan for Smart Water Resources,” have been introduced to promote the digital transformation of water management.

Extensive qualitative research has explored the potential impacts of this digital transformation on environmental governance. Most scholars agree that this digital transformation can enhance environmental governance [4,8,9,10]. On the one hand, the application of digital technologies, such as IoT, big data, and cloud computing, to environmental monitoring helps address issues like the lack of standardization in traditional monitoring data, difficulty in capturing emission outcomes, and the low precision of data monitoring [4,8]. These technologies enable real-time, accurate data collection and the comprehensive dynamic tracking of pollutants, thereby improving environmental decision-making and pollution control capabilities [9,10]. On the other hand, digital governance facilitates public access to environmental information and strengthens the supervisory roles of stakeholders, such as the public, insurance companies, and non-governmental organizations, thereby reducing the monopoly of polluters and governments over pollution information [10,11].

As research on the digital transformation of environmental governance has deepened, scholars have begun quantitative investigations to ascertain the actual effects of digital governance on pollution reduction. However, the findings remain inconsistent. For instance, Zhao et al. [12] found that the digital transformation of environmental governance contributes to pollution reduction and exhibits spatial spillover effects. Conversely, Kloppenburg et al. [10] argued that digital technologies do not automatically lead to better environmental outcomes or more democratic and inclusive governance methods. Based on the data from 41 OECD/EU countries between 2014 and 2020, Durkiewicz and Janowski [13] refuted the notion that digital governance inherently promotes sustainable development. Their empirical analysis revealed significant variations in the effectiveness of digital governance across countries, with some nations experiencing setbacks in sustainable development due to digital governance.

While the existing research recognizes the immense potential of digital governance for addressing environmental pollution, three gaps remain to be filled in the literature. Firstly, under the dual context of severe water pollution and the digital transformation of environmental governance, the effectiveness of digital governance for reducing water pollution requires further empirical validation. Secondly, as water pollution governance necessitates multi-stakeholder participation, particularly from manufacturing industries as the major contributors to industrial water pollution, there is a need to further explore how corporate technological innovation can adapt within the context of digital water management. Thirdly, the heterogeneous effects of digital governance on pollution outcomes, in terms of industry competitiveness and corporate ownership, remain an open question and need to be examined.

This study makes three primary contributions: (1) By dissecting the specific components of digital governance for water pollution, we categorize the tools into digital monitoring and digital administration, and analyze their respective impacts within a unified analytical framework. The findings reveal the effectiveness of digital tools for water pollution governance, offering new perspectives for understanding environmental governance in the digital age. (2) We examine the moderating effects of digital monitoring and digital administration on the relationship between corporate innovation and water pollution, uncovering the complex mechanisms through which different digital governance tools influence water pollution. This expands the scope of the research on the relationship between environmental regulation and corporate behavior. (3) We analyze the differential effects of digital water governance depending on industry competitiveness and corporate ownership types, providing more targeted insights for the formulation of corporate-level digital governance policies.

2. Theoretical Analysis and Research Hypotheses

2.1. Direct Effects of Digital Governance on Water Pollution

The direct impact of embedding digital technologies into water pollution governance is primarily reflected by two aspects: digital monitoring and digital administration.

Firstly, digital monitoring influences water pollution emissions through full-cycle tracking and precise early-warning. Specifically, digital technologies, such as sensors, the Internet of Things (IoT), and big data, enable the real-time and dynamic monitoring of the generation, transmission, and transformation of water pollution [14]. This full-cycle tracking ensures that pollution sources and changes are traceable, significantly improving monitoring accuracy and the timeliness of early warnings [1,15]. Consequently, it helps prevent water pollution incidents [15,16].

Secondly, digital administration transforms water pollution management from an inefficient, hierarchical coordination to a streamlined “one-to-many” collaborative model [17]. This approach breaks down the information silos between departments and regions [18,19], enhancing collaboration between government entities and between governmental and non-governmental organizations, thereby improving the efficiency of water pollution governance. Additionally, such a system sends water pollution issues directly to the respective responsible parties at all levels and automates follow-up and supervision [17]. This mechanism clarifies accountability, tracks task progress, and enables the timely resolution of water pollution issues, thereby enhancing the effectiveness of governance execution [17,18,19].

Based on the above analysis, we propose the following hypotheses:

Hypothesis 1: Digital monitoring reduces water pollution.

Hypothesis 2: Digital administration reduces water pollution.

2.2. Moderating Effects of Digital Governance on Water Pollution

The Porter Hypothesis posits that appropriately designed environmental regulations can stimulate corporate innovation, achieving a win–win outcome for both economic and environmental goals [20]. This theory provides a critical framework for understanding the relationship between the digital governance of water pollution and corporate behavior. Specifically, digital monitoring and digital administration influence the strength or direction of the relationship between technological innovation and corporate water pollution through mechanisms of “pollution data visibility” and “reputation enhancement,” respectively. On the one hand, digital monitoring enables the precise and timely tracing of pollution sources, facilitates the immediate handling of violations, shortens regulatory response times, and strengthens regulatory deterrence [21]. Under stricter regulatory environments, firms are compelled to enhance the quality of their technological innovations, improving their resource utilization efficiency and reducing their energy consumption to meet environmental standards and enhance their market competitiveness [22,23,24]. On the other hand, digital administration increases the transparency of corporate environmental behavior through multidimensional environmental information disclosure [25]. The heightened attention from the government, non-governmental organizations, the media, and the public compels enterprises to develop green technologies and enhance their environmental compliance to safeguard their reputations [26,27,28], thereby reducing water pollution.

Accordingly, we propose the following hypotheses:

Hypothesis 3: Digital monitoring positively moderates the relationship between technological innovation and water pollution.

Hypothesis 4: Digital administration positively moderates the relationship between technological innovation and water pollution.

3. Model Construction, Variable Selection, and Data Sources

3.1. Model Construction

3.1.1. Panel Fixed-Effects Model

Following Zhao et al. [29] and Xiong et al. [30], we employ a fixed-effects model to analyze the direct impact of digital governance on water pollution. The specific estimation model is as follows:

$${\mathrm{Wstwt}}_{\mathrm{qt}}{=\mathsf{\alpha }}_0{+\mathsf{\alpha }}_1\mathrm{Digital}\_{\mathrm{Mon}}_{\mathrm{pt}}{+\mathsf{\alpha }}_2\mathrm{Digital}\_{\mathrm{Adm}}_{\mathrm{pt}}{+\mathsf{\alpha }}_3{\mathrm{X}}_{\mathrm{qt}}+{\mathsf{\mu }}_{\mathrm{p} }+{\mathsf{\vartheta }}_{\mathrm{i}}$$

(1)

where ${ \mathrm{Wstwt}}_{\mathrm{qt}}$ represents corporate water pollution.$\mathrm{Digital}\_{\mathrm{Mon}}_{\mathrm{pt}}$ and $\mathrm{Digital}\_{\mathrm{Adm}}_{\mathrm{pt}}$ denote the levels of local government digital monitoring and digital administration, respectively. ${\mathrm{X}}_{\mathrm{qt}}$ represents control variables, including firm size (Size), leverage (Lev), profitability (ROE), R&D intensity (RD), and environmental protection intensity (Ptec). ${\mathsf{\mu }}_{\mathrm{p}}$ and ${\vartheta }_{\mathrm{i}}$ represent province fixed effects and industry fixed effects, respectively.

The digital governance variables are calculated using the entropy weight method, with 2015 as the base year. This approach partially controls for time trends. To avoid over-controlling, year fixed effects are not included in the baseline regression, as doing so may reduce the variability in explanatory variables.

3.1.2. Moderating Effects Model

To test the moderating effects of local government digital monitoring and digital administration on the relationship between technological innovation and water pollution, we construct the following model, following the approach of Xiong et al. [30]:

$$\begin{array}{c}{\mathrm{Wstwt}}_{\mathrm{qt}}{=\mathsf{\alpha }}_0{+\mathsf{\alpha }}_1\mathrm{Digital}\_{\mathrm{Mon}}_{\mathrm{pt}}\times {\mathrm{RD}}_{\mathrm{qt}}+{\mathsf{\alpha }}_2\mathrm{Digital}\_{\mathrm{Adm}}_{\mathrm{pt}}\times {\mathrm{RD}}_{\mathrm{qt}}+\\ {\mathsf{\alpha }}_3\mathrm{Digital}\_{\mathrm{Mon}}_{\mathrm{pt}} {+\mathsf{\alpha }}_4\mathrm{Digital}\_{\mathrm{Adm}}_{\mathrm{pt}}+{\mathsf{\alpha }}_5{\mathrm{RD}}_{\mathrm{qt}}{+\mathsf{\alpha }}_6{\mathrm{X}}_{\mathrm{qt}}+{\mathsf{\mu }}_{\mathrm{p}}+{\vartheta }_{\mathrm{i}}\end{array}$$

(2)

where $\mathrm{Digital}\_{\mathrm{Mon}}_{\mathrm{pt}}\times {\mathrm{RD}}_{\mathrm{qt}}$ and $\mathrm{Digital}\_{\mathrm{Adm}}_{\mathrm{pt}}\times {\mathrm{RD}}_{\mathrm{qt}},$ respectively, represent the interaction term of local government digital monitoring and digital administration with corporate technological innovation. These are the key variables for testing Hypotheses 3 and 4. Other variables are consistent with those in Model (1).

3.2. Variable Selection

3.2.1. Core Independent Variables

Currently, there is a paucity of studies on measuring the level of environmental digital governance. Zhao et al. [12] constructed a comprehensive index of digital environmental governance based on three dimensions: data system development, government administrative services, and digital monitoring. Similarly, Wen et al. [31] developed an index encompassing digital infrastructure, climate monitoring facilities, and climate governance investments to evaluate the digital climate governance levels across provinces. Based on these studies, and considering the processes and functionalities enabled by digital technologies for water pollution governance, we identify digital monitoring as a key component of digital governance of water pollution. Its primary role lies in transforming otherwise unobservable instances of pollution into real-time, accurate, and quantifiable data that can be easily understood [14,32]. The other key component of digital governance is digital administration, which enhances cross-departmental and multi-stage coordination in water pollution governance, improving the scientific rigor of policy-making and the timeliness of pollution management [33,34].

Therefore, we measure the digital governance of water pollution from two aspects: digital monitoring (Digital_Mon) and digital administration (Digital_Adm). The index system for these two variables is summarized in

Table 1. The index system for digital monitoring and digital administration.

Variable	Measurement Method
Digital Monitoring	Number of environmental monitoring instruments per unit of industrial added value.
	Number of surface water monitoring sites per unit area.
	Number of drinking water monitoring sites per unit area.
Digital Administration	Completeness of online government service methods.
	Coverage of online government service items.
	Accuracy of online government service guidelines.

.

In 2015, during the Second World Internet Conference, President Xi Jinping proposed the concept of “Digital China,” marking the start of rapid digital construction in the country. This initiative also established the foundation for measuring and obtaining data related to digital monitoring and digital administration of water pollution governance. Based on this context, 2015 is taken as the base year for calculating the level of digital water pollution governance. We calculate the digital monitoring and digital administration indicator using the entropy weight method with a fixed base, following Song et al. [35] and Zhou and Wu [36], as it objectively and accurately reflects changes in digital water pollution governance over the research period compared to the baseline year. The calculation process is as follows:

First, calculate the proportion of indicator j in year t for province p:

$${\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}={\mathrm{X}}_{\mathrm{pj}}^{\mathrm{t}}/{\displaystyle \sum }_{\mathrm{p}=1}^{\mathrm{n}}{\mathrm{X}}_{\mathrm{pj}}^{\mathrm{t}},\mathrm{p}\in \left[1,\mathrm{n}\right], \mathrm{j}\in \left[1,\mathrm{m}\right]$$

(3)

If ${\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}$ = 0, $\underset{{\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}\to 0}{\lim }{\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}\times \ln \left({\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}\right)$ = 0 is defined as 0, and ${\mathrm{X}}_{\mathrm{pj}}^{\mathrm{t}}$ represents the data that have been normalized using range standardization.

Next, calculate the information entropy of each indicator for year t:

$${\mathrm{E}}_{\mathrm{j}}^{\mathrm{t}}=-{\left[\ln \left(\mathrm{n}\right)\right]}^{-1}\times {\displaystyle \sum }_{\mathrm{p}=1}^{\mathrm{n}}[{\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}\times \ln \left({\mathrm{S}}_{\mathrm{pj}}^{\mathrm{t}}\right)]$$

(4)

where ${\mathrm{E}}_{\mathrm{j}}^{\mathrm{t}}$∈ [0, 1]; the smaller the information entropy of indicator j is, the greater the dispersion of j, which means it provides more information and consequently has a higher weight.

Then, calculate the weight of each indicator:

$${\mathrm{W}}_{\mathrm{j}}^{\mathrm{t}}=(1-{\mathrm{E}}_{\mathrm{j}}^{\mathrm{t}})/{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{m}}(1-{\mathrm{E}}_{\mathrm{j}}^{\mathrm{t}})$$

(5)

Then, using the initial year of the study period, 2015, as the baseline year, dimensionless normalization of the raw data is performed:

$${\mathrm{X}}_{\mathrm{j}}^{\mathrm{t}}=({\mathrm{x}}_{\mathrm{j}}^{\mathrm{t}}-{\mathrm{x}}_{\mathrm{j},\min }^{2015})/\left({\mathrm{x}}_{\mathrm{j},\max }^{2015}-{\mathrm{x}}_{\mathrm{j},\min }^{2015}\right)$$

(6)

where ${\mathrm{x}}_{\mathrm{j}}^{\mathrm{t}}$ represents the raw data, and ${\mathrm{x}}_{\mathrm{j},\max }^{2015}$ and ${\mathrm{x}}_{\mathrm{j},\min }^{2015}$ denote the maximum and minimum values, respectively, of indicator j for all provinces in the baseline year 2015.

Finally, calculate the comprehensive index of digital water pollution governance for province p in year t:

$${\mathrm{D}}_{\mathrm{p}}^{\mathrm{t}}={\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{m}}({\mathrm{W}}_{\mathrm{j}}^{\mathrm{t}}\times {\mathrm{X}}_{\mathrm{j}}^{\mathrm{t}})$$

(7)

3.2.2. Dependent Variable

Pollution emissions are a commonly used proxy indicator for pollution control performance [22,37]. Our corporate data are derived from the national tax survey database, which covers small and medium-sized enterprises, enabling a more comprehensive reflection of the impact of digital water pollution governance on corporate water pollution. However, this database only records the water pollution treatment fees paid by enterprises each year and does not provide data on annual water pollutant emissions. Therefore, we define the dependent variable, corporate water pollution (Wstwt), as the wastewater treatment fees paid by enterprises each year. This measurement aligns with the “polluter pays” principle, where enterprises that emit more wastewater are subject to higher treatment fees.

3.2.3. Control Variables

This study incorporates the following control variables: firm size (Size), leverage (Lev), profitability (ROE), R&D intensity (RD), and environmental protection investment intensity (Ptec). These variables are defined as follows. Firm Size (Size): measured by the total output value of an enterprise in a given year. Leverage (Lev): represents the corporate leverage ratio and is calculated as the ratio of total liabilities to total assets at the end of a year [38]. Profitability (ROE): measured as the ratio of operating profit to operating revenue. R&D Intensity (RD): measured as the ratio of R&D expenditures to operating revenue. Environmental Protection Intensity (Ptec): We used the ratio of the original value of environmental protection facilities to the volume of water pollutant emissions as a proxy. Given data limitations, the value of environmental protection facilities and emissions from the previous year are used.

3.3. Data Sources

This study’s provincial-level data from 2015 to 2020 came from China Ecological Environment Statistics Bulletin, China Environmental Statistical Yearbook, Provincial Government Online Government Service Capability Assessment Report, and China Statistical Yearbook. Firm-level data were sourced from the national tax survey database for 2015–2016. All data processing and analysis were conducted using Stata 17.

The sample was refined in the following ways: (1) Enterprises other than manufacturing enterprises were eliminated. (2) Firms with missing key variables (such as employee numbers, assets, total output, revenue, wastewater treatment fees, R&D expenditures, or prior-year environmental investments) were excluded. (3) Unreasonable data points (e.g., negative or zero revenues, wastewater treatment fees, or employee numbers) were removed. (4) Some of the original data for the asset–liability ratio and profitability contained negative values, which could not be directly log-transformed. Thus, we standardized these variables to a 0–1 range. (5) For other control variables and the dependent variable, we performed log transformation. To avoid undefined log values for observations with zero number, we replaced zeros with a near-zero value (0.0000000001).

The descriptive statistics for variables are shown in

Table 2. Descriptive statistics.

Variable	Obs	Mean	Std.dev.	Min	Max
Digital_Mon	67,441	0.113	0.075	−0.007	0.783
Digital_Adm	67,441	0.719	0.253	0.094	0.969
Wstwt	67,441	1.356	3.690	−4.605	13.703
Size	67,441	3.379	2.180	−6.908	12.965
Lev	67,441	0.564	0.274	0.000	1.000
ROE	67,441	0.595	0.005	0.000	1.000
RD	67,441	−18.212	10.164	−23.026	11.722
Ptec	67,441	−20.106	9.452	−23.026	30.290

.

3.4. Multicollinearity Test

In order to ensure the reliability of the regression results, we used Pearson correlation method to test for the presence of multicollinearity. When the correlation coefficient is close to or equal to ±1, there is multicollinearity problem. As shown in

Table 3. Multicollinearity test.

	Wstwt	Digital_Mon	Digital_Adm	Size	Lev	ROE	RD	Ptec
Wstwt	1
Digital_Mon	0.0140 ***	1
Digital_Adm	0.2304 ***	−0.1147 ***	1
Size	0.4180 ***	−0.0363 ***	0.2282 ***	1
Lev	0.0491 ***	0.0099 **	0.1232 ***	0.1065 ***	1
ROE	0.0138 ***	−0.0124 ***	0.0180 ***	0.0493 ***	0.0121 ***	1
RD	0.1774 ***	−0.0187 ***	0.0369 ***	0.2752 ***	−0.0770 ***	0.0003	1
Ptec	−0.0445 ***	−0.0039	0.0091 **	0.1430 ***	−0.0197 ***	0.0042	0.0767 ***	1

Notes: ** p < 0.05; *** p < 0.01.

, the correlation coefficients between independent variables are significantly less than ±1, indicating that there is no multicollinearity. This ensures the reliability of subsequent regression analysis.

4. Empirical Results and Discussion

4.1. Benchmark Regression Analysis

Table 4. Baseline regression results.

	(1)	(2)
Digital_Mon	2.773 ***	3.081 ***
	(7.36)	(8.59)
Digital_Adm	−6.883 ***	−7.286 ***
	(−42.41)	(−48.22)
Size		0.628 ***
		(94.16)
Lev		−0.224 ***
		(−5.03)
ROE		−4.242 *
		(−1.96)
RD		0.030 ***
		(22.14)
Ptec		−0.030 ***
		(−22.19)
_cons	5.990 ***	6.717 ***
	(42.05)	(5.18)
Province fixed	Yes	Yes
Industry fixed	Yes	Yes
N	67,441	67,441
R2	0.161	0.296

Notes: t statistics in parentheses; * p < 0.10; *** p < 0.01.

presents the regression results for the effects of digital monitoring (Digital_Mon) and digital administration (Digital_Adm) on corporate water pollution emissions. Column (1) shows the results without the control variables, while Column (2) incorporates the control variables.

Specifically, the coefficient of Digital_Mon is 3.081, which is statistically significant at the 1% level (Column (2) of Table 4). This indicates a significant positive correlation between government-led digital water pollution monitoring and corporate water pollution emissions, which does not align with Hypothesis 1. A possible explanation is that the primary function of digital monitoring is to accurately identify pollution sources, quantify the severity of pollution, and promptly detect pollution incidents. The application of digital monitoring technologies makes pollution data more transparent, improving identification efficiency and increasing the visibility of polluting behaviors [1,9,10,15,16]. Consequently, this transparency may result in higher reported pollution levels in the data.

In contrast, the coefficient of Digital_Adm is −7.286 and statistically significant at the 1% level (Column (2) of Table 4). This implies that, on average, a one-unit improvement in digital administration quality leads to a 7.286% reduction in corporate water pollution emissions. This finding is consistent with studies by Tang et al. [39] and Castro and Lopes [40]. Additionally, the result is aligned with Column (1) of Table 4, demonstrating the robustness of this study’s findings and providing strong support for Hypothesis 2.

4.2. Moderation Effects Analysis

To test Hypotheses 3 and 4, which suggest that Digital_Mon and Digital_Adm may encourage technological innovation to reduce pollution, we conducted a regression estimation on Model 2. The results are shown in

Table 5. Moderation effects analysis.

	(1)	(2)
Digital_Mon × RD	0.0003	−0.0076
	(0.01)	(−0.48)
Digital_Adm × RD	−0.0273 ***	−0.0357 ***
	(−5.01)	(−6.79)
RD	0.0476 ***	0.0571 ***
	(9.49)	(11.70)
Digital_Mon	2.2283 ***	2.8929 ***
	(6.20)	(6.41)
Digital_Adm	1.6416 ***	−7.9592 ***
	(14.00)	(−43.93)
Size	0.6430 ***	0.6247 ***
	(95.28)	(93.48)
Lev	−0.0917 **	−0.2317 ***
	(−1.98)	(−5.21)
ROE	−5.1627 **	−4.0544 *
	(−2.45)	(−1.84)
Ptec	−0.0415 ***	−0.0297 ***
	(−31.06)	(−22.22)
_cons	0.5539	7.1457 ***
	(0.44)	(5.40)
Province fixed	No	Yes
Industry fixed	No	Yes
N	67,441	67,441
R2	0.212	0.296

Notes: t statistics in parentheses; * p < 0.10; ** p < 0.05; *** p < 0.01.

. Column (1) excludes the fixed effects, while Column (2) includes the province and industry fixed effects.

The coefficient of RD is 0.0571, and reaches the statistical significance level of 1%, which means that the improvement in an enterprise’s technological innovation level aggravates water pollution (Table 5, Column (2)). It appears that technological innovation is a double-edged sword, which can not only save resources by improving productivity, but can also cause the expansion of the production scale of enterprises, thus resulting in more water pollution [37,41].

The coefficients of Digital_Mon × RD are statistically insignificant (Table 5, Column (2)), indicating an inconclusive moderating effect of government-led digital water pollution monitoring on the relationship between corporate innovation and water pollution. A potential reason for this is that digital monitoring technologies mainly operate as tools to assist and enhance traditional environmental regulatory methods. Without regulatory mechanisms that include incentives or penalties, environmental digital monitoring alone simply functions as a supplementary tool for observation and cannot independently achieve substantive governance effects.

In contrast, the coefficients of Digital_Adm × RD are negative and statistically significant at the 1% level (Table 5, Column (2)). This reveals that digital administration positively moderates the relationship between corporate innovation and water pollution, thereby supporting Hypothesis 4. As digital administration quality improves, the detrimental effects of corporate technological innovation on water pollution are mitigated. A plausible explanation for this is that a digital government platform connects corporate pollution control performance to financing and loan activities, incentivizing firms to focus on green technology and high-end innovation [42], thereby reducing water pollution.

4.3. Heterogeneity Analysis

4.3.1. Heterogeneity Analysis of Firm Ownership

The ownership structure of firms has a significant impact on their operational goals, decision-making mechanisms, and access to resources [43]. The effects of digital water pollution governance on water pollution may vary by firm ownership structure, resulting in heterogeneity. To explore this, we classified firms into state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) to examine whether the pollution-reduction effects of digital governance differ across ownership types.

The regression results indicate that the impact coefficient of Digital_Mon for SOEs is −3.247, which is not statistically significant (

Table 6. Heterogeneity analysis.

	(1)	(2)	(3)	(4)
	SOEs	Non-SOEs	Low Competition	High Competition
Digital_Mon	−3.247	3.316 ***	2.145 ***	4.381 ***
	(−1.48)	(9.11)	(4.62)	(7.67)
$\mathrm{Digital}\_\mathrm{Adm}$	−8.860 ***	−7.278 ***	−7.732 ***	−6.811 ***
	(−7.98)	(−47.70)	(−37.12)	(−30.92)
Size	0.549 ***	0.629 ***	0.648 ***	0.600 ***
	(12.37)	(92.97)	(68.80)	(63.94)
Lev	0.077	−0.229 ***	−0.190 ***	−0.255 ***
	(0.23)	(−5.10)	(−2.96)	(−4.14)
ROE	−299.898	−4.208 *	−1.923	−7.737 **
	(−1.33)	(−1.94)	(−0.80)	(−2.38)
RD	0.063 ***	0.029 ***	0.032 ***	0.027 ***
	(5.41)	(21.60)	(17.66)	(13.48)
Ptec	−0.019 *	−0.030 ***	−0.029 ***	−0.031 ***
	(−1.68)	(−22.17)	(−15.67)	(−15.75)
_cons	184.089	6.669 ***	5.718 ***	8.350 ***
	(1.37)	(5.14)	(3.95)	(4.29)
Province fixed	Yes	Yes	Yes	Yes
Industry fixed	Yes	Yes	Yes	Yes
N	1072	66,368	33,730	33,711
R2	0.374	0.296	0.292	0.298

Notes: t statistics in parentheses; * p < 0.10; ** p < 0.05; *** p < 0.01.

, Column (1)). However, for non-SOEs, the coefficient of Digital_Mon is 3.316, and it is statistically significant at the 1% level (Table 6, Column (2)). This suggests that digital water pollution monitoring is significantly positively correlated with the water pollution emissions of non-SOEs, while it has no significant effect on that of SOEs.

The regression coefficient of Digital_Adm for SOEs is −8.860 (Table 6, Column (1)), and for non-SOEs, it is −7.278 (Table 6, Column (2)). Both coefficients are statistically significant at the 1% level. This indicates that digital administration significantly reduces water pollution emissions of both SOEs and non-SOEs, but the effect is more pronounced for SOEs.

One possible explanation is the dual nature of SOEs’ value goals, combining profit-making and public welfare. When faced with environmental regulations, SOEs tend to focus more on compliance with environmental quality standards during production while increasing green investments to achieve a balance between economic and environmental performance [44,45]. Consequently, SOEs do not report higher pollution levels despite adopting digital water pollution monitoring technologies. Furthermore, as digital administration improves, SOEs are more likely to comply with policy guidelines and undertake proactive measures to reduce water pollution.

4.3.2. Heterogeneity Analysis of Industry Competition

Industry competition exerts a substantial influence on firms’ competitive strategies. The effect of digital governance on water pollution may vary based on the level of competitive pressure faced by firms. Following Haushalter et al. [46] and He et al. [47], we measure industry competition using the Herfindahl–Hirschman Index (HHI). Specifically, the HHI = $\sum {\left({\mathrm{T}}_{\mathrm{qi}}/\sum {\mathrm{T}}_{\mathrm{i}}\right)}^2$, where ${\mathrm{T}}_{\mathrm{qi}}$ represents the main business income of enterprise q in industry i, and $\sum {\mathrm{T}}_{\mathrm{i}}$ represents the main business income of all enterprises in industry i. A lower HHI value indicates greater industry competition, implying higher competitive pressure. We categorize industries into two groups based on the median HHI. Industries at or below the median are classified as highly competitive, while those above the median represent less competitive industries. Separate regressions are conducted for the firms operating in highly competitive and less competitive industries (Table 6).

The results show that the coefficient of Digital_Mon is 2.145 for low-competition industries and 4.381 for high-competition industries, both statistically significant at the 1% level (Table 6, Columns (3) and (4)). Conversely, the coefficients of Digital_Adm are −7.732 for low-competition industries and −6.811 for high-competition industries, also significant at the 1% level (Table 6, Columns (3) and (4)).

These findings suggest that Digital_Mon has a stronger role in aggravating the water pollution emissions of firms in high-competition industries. Meanwhile, Digital_Adm more effectively suppresses water pollution in low-competition industries. A plausible explanation for this is that firms in highly competitive industries face significant survival pressures and prioritize market share and short-term profits over environmental responsibility [48,49]. This focus on immediate gains diminishes the effectiveness of digital water pollution governance in high-competition industries.

4.4. Robustness Analysis

We adopt two approaches to examine the robustness of the results. First, since the baseline regression results are derived from a micro-level analysis of firm data, we perform a cross-validation using 2015–2020 macro-level provincial data. Following the STIRPAT model, which is commonly used to study factors influencing regional environmental quality [37,50,51], we include population, affluence, and technology as the variables. Additionally, we control for industrial structure and openness. Specifically, considering the availability and representativeness of the data, we use Zhou et al. [50] and Fan and Fang [52] as references for measuring water pollution with the logarithm of industrial COD emissions. The population is represented as the logarithm of the resident population. Affluence is measured by the logarithm of per capita GDP. Technology is proxied for by the logarithm of science and technology spending as a percentage of the GDP. The industrial structure is captured by the logarithm of the tertiary industry’s share of the GDP. Openness is represented by the logarithm of trade volume as a percentage of the GDP. The results (Column (1) of

Table 7. Robustness tests.

	(1)	(2)
Digital_Mon	1.209 **	3.081 ***
	(2.46)	(8.33)
Digital_Adm	−0.508 *	−7.286 ***
	(−1.75)	(−48.12)
Control variables	Yes	Yes
Province fixed	Yes	Yes
Industry fixed		Yes
_cons	52.682 ***	6.717 **
	(3.02)	(2.56)
N	180	67,441
R2	0.667	0.296

Notes: t statistics in parentheses; * p < 0.10; ** p < 0.05; *** p < 0.01.

) indicate significant coefficients of 1.209 for Digital_Mon and −0.508 for Digital_Adm, consistent with the baseline regression results. This suggests that digital water pollution monitoring significantly correlates with higher pollution levels, while digital administration significantly reduces emissions.

Second, since the bootstrap method does not rely on data distribution assumptions and provides reliable standard error estimates through repeated sampling [53], we use the bootstrap method to conduct 500 repeated samplings and re-estimate the model with the regression results shown in Column (2) of Table 7. The estimation results are basically consistent with the benchmark regression results, indicating again that the conclusions of this study are robust.

5. Conclusions, Policy Implications, and Limitations

5.1. Research Conclusions

This study employs a fixed-effects model to explore the impact of government-led digital water pollution governance on the water pollution of manufacturing enterprises. The baseline regression results reveal an asymmetric effect of Digital_Mon and Digital_Adm on corporate water pollution. Specifically, Digital_Mon is significantly positively correlated with corporate water pollution emissions, indicating that the application of digital monitoring technologies enhances the detection rate of water pollution but lacks the regulatory capacity to effectively reduce emissions on its own. In contrast, Digital_Adm significantly reduces corporate water pollution emissions, suggesting that digital administration promotes interdepartmental collaboration regarding water pollution governance, effectively driving enterprises to reduce pollution. The moderation analysis shows that digital administration positively moderates the relationship between corporate technological innovation and water pollution, weakening the negative impact of innovation on pollution. However, the moderating effect of digital monitoring is not significant, highlighting that the regulatory effectiveness of digital monitoring depends on its integration with other regulatory mechanisms. The heterogeneity analysis indicates that non-state-owned enterprises (non-SOEs) and firms facing higher industry competition are more sensitive to digital water pollution monitoring. Conversely, the pollution-reduction effect of digital administration is stronger among state-owned enterprises (SOEs) and firms operating in less competitive industries.

5.2. Policy Implications

To enhance the effectiveness of digital environmental monitoring regulations, unified water quality monitoring parameters and equipment standards need to be established to ensure the continuity and comparability of water quality data, providing a reliable basis for subsequent analyses of water quality issues. As water pollution monitoring capabilities improve, higher demands will also be placed on the human resources for addressing water pollution. On the one hand, when water pollution is detected in a specific location, staff will still be needed to verify and handle the pollution on-site, which will significantly increase the workload of grassroots workers. On the other hand, the volume of monitoring data will increase sharply with higher monitoring frequencies, posing a challenge to government staff’s data analysis capabilities. Therefore, government departments need to not only update and apply advanced water quality monitoring technologies but also increase the number of data analysis professionals and grassroots environmental law enforcement officers. By promoting the full-chain coordination of water pollution data collection, analysis, and law enforcement, the effectiveness of water pollution control can be improved.

To strengthen the role of digital administration in environmental regulations, streamlined business processes should be established based on the types of businesses and the severity of the pollution problems involved. This would promote data coordination among government departments regarding corporate pollution and environmental credit, reducing the administrative burdens on enterprises and enhancing the efficiency of environmental governance. In addition, digital administration platforms should promptly publish green development incentive lists to align the information that enterprises and the government have on green incentives, encouraging enterprises’ sustainable development. Further, the public could be better informed about environmental issues and their awareness of pollution could be improved by strictly implementing environmental information disclosure systems and adding environmental knowledge learning modules to digital administration platforms. This multi-faceted supervision could pressure enterprises to prioritize environmental impacts. Lastly, to continuously optimize digital environmental administration services, governments could focus on satisfaction surveys and ensure that the feedback from these surveys is promptly addressed, thereby increasing the participation of enterprises and the public in environmental protection.

Given the heterogeneous effects of digital water pollution monitoring and administration on pollution, in terms of firm ownership and industry competitiveness, differentiated environmental regulations should be developed for enterprises of different types. For non-state-owned enterprises and those facing intense industry competition, governments should strengthen the digital monitoring of water pollution by mandating the installation of real-time-feedback pollution monitoring equipment to curb opportunistic behavior. At the same time, the relevant government departments could release low-cost green technology lists and organize related seminars to assist enterprises facing high competition, low profits, and large output with how to reduce the costs of green transformation and achieve gradual green upgrades. For state-owned enterprises and those in less competitive industries, governments could provide incentives, such as R&D subsidies and tax preferences, to encourage more proactive environmental responsibility fulfillment.

5.3. Limitations and Future Research Directions

Firstly, due to the unavailability of environmental digital monitoring data below the provincial level, this study shares a common limitation with existing research on environmental digital governance, that is, the measurement of digital water pollution governance remains confined to the provincial level. Additionally, the disclosure of environmental digital monitoring and digital administration data at the provincial level began relatively recently, resulting in a limited study period. As data disclosure continues to expand, future research could assess the practical effects of digital water governance by expanding the time span of the data and delving into more granular data levels.

Secondly, this study employs quantitative methods to evaluate the effects of government-led digital water pollution monitoring and digital administration on corporate water pollution. Future studies could complement the current study with qualitative approaches, such as interviews and surveys of government officials and corporate staff across different cities, to uncover the underlying micro-mechanisms. Such efforts would provide deeper insights and further enrich the research on environmental digital governance.