How digital finance affects environmental pollution management: evidence from China

Due to people’s insufficient anticipation of the negative impact of highly developed industries and the lack of prevention, global environmental pollution has occurred. These pollutants include air pollution, water pollutants, and land pollution, which not only cause direct damage and impact the ecosystem but also endanger the health of urban residents and economic development. Therefore, researching environmental pollution management is necessary to help solve these imminent environmental problems. In addition, digital finance, based on digital technology, can identify bottlenecks in environmental pollution management, formulate more effective governance strategies, and reduce environmental pollution at the source. In this context, this study uses the environmental pollution data of 287 cities in China from 2011 to 2021. It uses the fixed-effects and mediation effect models to analyze digital finance’s role in environmental pollution management. The research shows that digital finance can promote environmental pollution management and play a promoting role through two channels of influence: green technology innovation and government green subsidies. At the same time, the effect of this promotion is more significant in cities in the Midwest and in resource-based cities. The research results propose strategies for government organizations in environmental pollution management, and alleviate current resource and environmental problems, in addition to realizing sustainable urban development.


Introduction
The global economy has developed rapidly over the past few decades. Along with accelerating industrialization and urbanization, economic growth has been impacting the ecological environment significantly and negatively (Başar and Tosun 2021). The development of modern industries has brought about a large amount of resource and energy consumption, leading to increased pollution of air, water, soil, and other environments. Discharging industrial wastewater, waste gas, and solid waste pollutes water sources and soil, and the discharge of radioactive substances endangers the environment and human health (Liang et al. 2019). Taking China as an example, according to the report of the Ministry of Ecology and Environment of China, in 2021, among the 331 prefecture-level cities in China, the number of cities whose ambient air quality meets the standard accounts for 64.3% of the total number of cities, and half of the cities in the country still have air quality problems. In terms of water pollution, the survey results of the "2023 China Modern Ecological Development Index" show that 59.3% of the respondents rated the water quality of daily water as "average," and 13.4% of the respondents considered it "relatively poor." Only 24.7% of the respondents gave positive comments, which is less than a quarter of the respondents. In the long run, environmental pollution will lead to high costs of resource loss and environmental restoration, which seriously restricts the healthy development of the economy. Improving the level of environmental pollution management has become an essential task at the current stage of development.
The literature on the influencing factors of environmental pollution management focuses on foreign investment (Lan et al. 2012;Sapkota and Bastola 2017), industrial structure (Liu and Lin 2019), and environmental regulations (Feng et al. 2020). However, these influence methods focus on restricting the discharge of environmental pollution at the end of the production process, overlooking the restrictions on pollution sources. While digital finance can use digital technology to realize process digitization and paperless processes, it also improves the efficiency of server resource usage, reduces power consumption, and achieves low emissions through a distributed architecture. Utilizing financial means restricts the financing, payment, and investment of target companies and further helps cities achieve green and sustainable development (Zhong 2022). However, there is scant empirical evidence on how digital finance impacts environmental pollution management. Can digital finance promote urban environmental pollution management? Through what channels does digital finance influence environmental pollution management? Does the effect of digital finance on environmental pollution management exist in the heterogeneity of city types?
To solve these problems, we selected panel data from 287 cities in China from 2011 to 2021 to explore whether digital finance can improve environmental pollution management and analyze the intermediary role of green technology innovation and government green subsidies. Our marginal contributions mainly include the following three parts: First, the literature mainly focuses on environmental analysis at the macro level , with little attention paid to the city level. Cities are usually the concentration of pollution problems, and research at the city level can locate the problem more accurately. Therefore, this study is different from the existing literature, analyzing the impact of digital finance on environmental pollution management from the city level. It reveals the changes in China's urban environmental pollution governance in more detail. The research conclusions can directly guide urban planning and pollution source management. It also helps the government and related agencies formulate targeted policies and measures. Second, most of the current research on environmental pollution focuses on carbon dioxide pollution (Lei et al. 2023) and rarely considers other pollutants. We have established an environmental pollution management indicator system covering sewage, waste gas, and smoke particles. Through visualization, the system can directly display the level of environmental pollution management in Chinese cities, highlight distribution differences, and make precise formulations in the decision-making and planning process. Third, the literature analysis on digital finance and environmental pollution impact channels mainly focuses on industrial upgrading and financing constraints. These two aspects are based on the subjective perspective of enterprises and industries. Market mechanisms alone may not be able to fully achieve the goal of reducing environmental pollution. It needs to be considered from the perspective of the government. Therefore, this study explores the impact path from the perspective of the government and analyzes the mediating role of government green subsidies and green technology innovation. Such analysis helps to broaden the research scope of environmental management and has important practical significance on reducing urban environmental pollution and realizing sustainable regional development.
The rest of this paper is organized as follows: the "Literature review" section provides an overview of related research; the "Research hypotheses" section conducts theoretical analysis and proposes relevant hypotheses; the "Methodology and data sources" section describes the data and methods; the "Statistical measurement and empirical analysis" section reports the benchmark regression results and robustness analysis; the "Mediation effect and heterogeneity analysis" section conducts an empirical analysis on the heterogeneity of the influencing channels. The "Conclusion and suggestion" section presents the conclusions of this research and makes recommendations.

Digital finance
There are many studies on digital finance, mainly focusing on the connotation of digital finance and its influencing factors. In the related research on the connotation of digital finance, some scholars define digital finance as a model in which traditional financial institutions and Internet companies use digital technology to realize financing, payment, investment, and other new financial services (Feng et al. 2022). According to the characteristics of digital finance, some scholars believe that digital finance embodies subversive innovations in economic activities such as credit intermediation, information collection, risk pricing, and investment decision-making ). Definitions of digital finance have not yet converged. Nevertheless, there is some consensus that digital finance is financial services provided via mobile phones, computers, the Internet, or cards linked to a reliable digital payment system (Ozili 2018). In addition, through combing the literature, we found that the research on the impact of digital finance can be divided into two aspects: macro and micro.
On the one hand, digital finance helps promote crossborder trade and investment, boosting economic growth (Sun and Tang 2022). It has also been shown that digital finance significantly mitigates economic imbalances between cities by leveraging its inclusive characteristics and that its impact is particularly pronounced in rural areas (Wang 2023). At the same time, the development of digital finance has significantly promoted the upgrading of the city's industrial structure and has a positive spatial spillover effect. The effect of promoting funds on industrial structure upgrading is more significant in cities with more developed economies, lower degrees of financialization, and lower income inequality (Ren et al. 2023). Specific to the manufacturing level, digital finance has a significant positive impact on the servitization of the manufacturing industry, and it promotes the servitization of the manufacturing industry by increasing the intensity of innovation and improving the level of digitalization. The role of digital finance conforms to the law of diminishing marginal utility, and on this basis, an obvious superimposed effect has been formed (Chen and Zhang 2021).
On the other hand, from the perspective of microcosmic household consumption, digital finance can promote the upgrading of household consumption (Hu et al. 2023). The promotion effect of digital finance on consumption is more obvious in families with fewer assets, lower income, and poorer financial literacy. It has greatly increased the number of women in the family-bargaining power (Han et al. 2023). In addition, digital finance mostly focuses on promoting recurring consumption, as online shopping has become an important channel for digital finance in terms of affecting household consumption (Li et al. 2020). Meanwhile, digital finance hurts household consumption. It has broadly increased the participation of residents in the credit market, increased the marginal propensity to consume outside of household economic mobility, and increased household financial risks (Yue et al. 2022).

Digital finance and environmental pollution
With the increasing seriousness of the global environmental pollution problem, the research content of ecological pollution control has gradually been enriched. Upon reviewing the literature, it can be observed that the research on digital finance and environmental pollution mainly focuses on carbon emissions. Chu et al. (2023) found that digital finance can suppress carbon emissions through a study on China's county-level carbon emissions, and this inhibitory effect is more significant under stronger financial and environmental regulation. Lee and Wang (2022) proposed that digital finance's inhibitory effect on carbon emissions has prerequisites such as economic scale and degree of openness. At the same time, digital finance has a spatial spillover effect on carbon emission suppression, and the green credit policy is most effective in suppressing carbon emissions within 350 to 400 km of a city. Wang and Guo (2022) proposed that digital finance can effectively reduce carbon emissions in household consumption and can lend more focus to smalland medium-sized vulnerable groups ignored by traditional financial institutions to improve fund availability. This inclusive function will encourage enterprises and individual investors to invest in green technology investments, thereby curbing carbon emissions at the source. At the same time, some scholars believe that the digital divide caused by digital finance has exacerbated residents' carbon emissions and caused changes in the carbon emission structure (Zhengning and Jinhua 2022).
In summary, the literature has researched the two links between digital finance and environmental pollution, but the following deficiencies exist. First, most current research considers environmental pollution from the perspective of carbon emissions while overlooking other industrial pollutant emissions, and the measurement method of environmental pollution is too simple. Second, the relationship between digital finance and environmental pollution has not reached a unified consensus. Third, the impact paths of digital finance on environmental pollution are concentrated on industrial structure (Du et al. 2022) and investment in research and development (Ma et al. 2022), with little analysis of other impact paths. Based on the above analysis, we use wastewater, waste gas, and smoke particles to comprehensively construct an environmental pollution index and study the role of digital finance in environmental pollution management. In addition, this study analyses the impact path from the perspectives of green technology innovation and government green subsidies. It analyzes the heterogeneity between resource-based cities and nonresource cities. It enriches the current research and provides a reference for environmental pollution management and urban green development.

Direct impact of digital finance on environmental pollution management
Digital finance combines digital technology with traditional financial services to form digital financial services through artificial intelligence, cloud computing, and other information technology means (Razzaq et al. 2023). From a theoretical analysis perspective, the impact of digital finance on environmental pollution management mainly includes the following three aspects: First, digital financial services can support the rapid and accurate collection and monitoring of environmental data. Technologies such as sensors and IoT devices can monitor environmental parameters such as air quality, water quality, and energy consumption in real time, thereby providing real-time pollution information and helping detect pollution sources and abnormal situations early. At the same time, digital finance has the characteristics of inclusiveness, which can broadly improve the public's environmental awareness and participation in green undertakings and promote the development of green public utilities in the urban field. The mobile payment method dominated by the digital payment platform can publicize and popularize environmental protection knowledge among consumers through the Internet, generate knowledge spillover effects, and enhance the awareness of urban green ecological protection.
Second, digital finance has the characteristics of digitization and informatization. These characteristics can enrich the paradigm of financial products, improve the level and efficiency of financial services, overcome the disadvantages of traditional financial information circulation channels, and enhance the transparency of financial market information. Next, financial institutions can use digital technology to accurately grasp the financial needs of SMEs, provide stable financial support for SMEs, and promote the digital transformation of SMEs. Reduce the dependence of enterprises on natural resources in terms of the production process and contribute to the green development of cities .
Finally, digital finance uses financial tools to guide the flow of social funds and optimize the allocation of resources. Through macroeconomic policies, the government guides funds to low-pollution and environmentally friendly enterprises, helps green industries alleviate complex financing problems, and uses policy dividends to encourage enterprises toward green transformation. At the same time, digital finance can use financial tools to monitor the financial market effectively, reasonably predict market risks and environmental risks in financial business transactions, and establish corresponding hedging mechanisms to deal with environmental threats and external systemic risks in urban development. In addition, digital finance can play a regulatory role, such as raising financial constraints on high-polluting enterprises by raising loan thresholds and restricting listings. However, digital finance can impose financial restrictions on high-polluting and high-emission enterprises by raising credit thresholds, restricting financing, and strengthening environmental information disclosure supervision. While effectively improving the level of environmental pollution control, it can reduce the credit risk of the financial market. Therefore, it fulfills its regulatory function. Based on the above analysis, we propose the following hypothesis: Hypothesis 1: Digital finance can promote environmental pollution management.

The mediating role of green technology innovation and government green subsidies
Upon combing the literature, we found that existing research has focused on the impact mechanism of digital financial development on environmental pollution management from the perspective of industrial structure and other paths (Du et al. 2022). However, little literature exists on the two paths of green technological innovation and government green subsidies. From the perspective of green technology innovation, digital financial platforms provide data analysis and management tools to facilitate green technology innovation. Data analysis and artificial intelligence technology can process large amounts of environmental data and collect valuable information. These data can identify opportunities for green technology innovation, assess their potential impact, and provide decision support to innovators, thus accelerating the R&D process. On the other hand, green digital technology has enormous advantages in analyzing environmental data. Green technology innovation is conducive to the development of the clean energy industry. Only by effectively preventing and reducing environmental pollution from the front end of the supply chain, optimizing production methods, strengthening the use of green energy, and increasing investment in pollution control equipment can large pollution emissions be fundamentally alleviated and environmental pollution management be improved. Based on this, we propose the following hypothesis: Hypothesis 2: Digital finance can promote environmental pollution management by promoting innovation in green technologies.
In the wave of global green transformation, government green subsidies can help domestic companies gain competitive advantages in the green field (Hussain et al. 2022). Green subsidies can improve environmental quality by reducing companies' energy emissions (Li et al. 2018). Specifically, on the one hand, traditional subsidy allocation may have problems such as information asymmetry and cumbersome procedures, resulting in a waste of resources and ineffective results. Digital financial technology can establish an intelligent subsidy management system to realize automated application, review, and issuance processes; reduce labor costs and time; and improve the timeliness and accuracy of subsidies. On the other hand, companies usually face cost pressures in terms of investment in environmental protection, and green subsidies can partially offset the cost increase brought by environmental protection investment, making cleaner production more attractive. Companies will be more inclined to adopt environmental protection measures when they are fighting for green subsidies, which will help reduce the pollution of urban air and water bodies and improve citizens' living environment. Therefore, based on the above theoretical analysis, we propose the following hypothesis: Hypothesis 3: Digital finance can promote environmental pollution management by increasing government green subsidies.
We propose here a road map for the above assumptions, as shown in Fig. 1.

Data sources
To analyze the impact of digital finance on environmental pollution management, we use Chinese cities from 2011 to 2021 as the research object. Considering that some cities lack key data, we eliminate such cities. After screening, we obtained 287 cities as research samples. In terms of data sources, the data for this study come from the Peking University Institute of Digital Finance, China Statistical Yearbook, and China Fiscal Statistical Yearbook. At the same time, to exclude the influence of outliers, this study carried out 1% shrinkage processing on continuous variables.

Dependent variable
Environmental pollution management (POL). Environmental pollution management is a comprehensive management method aimed at reducing, preventing, or eliminating environmental pollution and protecting natural ecosystems and human health. Its core is to control the emission of pollution sources and reduce the release of harmful substances and pollutants to the environment. In existing studies, most investments in environmental pollution management are measured by carbon dioxide emissions (Azam et al. 2018;Sohail et al. 2021). However, this research believes that using carbon emissions to measure environmental pollution is relatively simple and that other industrial and domestic pollution, such as sewage, has not been considered. Therefore, to consider the types of pollutants more comprehensively, we use the three wastes to construct the environmental pollution management level index. The measurement method of the three waste types, including the total discharge of urban sewage, sulfur dioxide discharge, and soot discharge, has been confirmed in existing studies ). Then, this study uses the entropy method to measure the environmental pollution management level of 287 sample cities. In addition, the environmental pollution indicators showed positive performance in the total indicators after standardized processing. The higher the calculation result of the entropy weight method is, the higher the level of environmental pollution management in the city. The specific index system and the calculation process of the entropy value are shown in Table 1.
Step 2: Calculate the proportion of the i-th sample value under the j-th index to the index.
(1)  Step 3: Calculate the entropy value and difference coefficient of the j-th indicator.
Step 4: Calculate the weight of the index j.
Step 5: Multiply the weight and the standardized value correspondingly to obtain the environmental pollution control level score.
Max (X ij ) is the maximum value of each year's index value, min(X ij ) is the minimum value of each year's index value, and X ij is the standardized index value. Due to space constraints, we show the results of the environmental pollution management level calculations of 31 provincial capital cities in the "Appendix" section.

Independent variable
Digital finance (DIFI). Digital finance refers to how digital and information and communication technologies are used to improve, enhance, and innovate financial services. We draw on research related to digital finance (Guo et al, 2020). The digital financial inclusion index released by the Peking University is selected to measure the level of urban digital finance. In addition, to analyze the impact of digital finance on environmental pollution management from multiple perspectives, we selected three indicators: degree of digitalization (dig), depth of use (use), and coverage (cov) to improve the accuracy of this research.

Mediator variable
(1) Green technology innovation (GTI). As global environmental problems continue to intensify and people pay increasing attention to climate issues, the importance of green technology innovation has become increasingly prominent. In the literature, the number of patent applications is often used to measure technological innovation capability. To distinguish green patents from other types of patents, we draw on the practice of , which uses the patent classification code (IPC) of the State Intellectual Property Office to classify patents and select green patents to measure green technological innovation. To ensure the scientificity and reliability of the research, we add 1 to the natural number of green invention patents and take the logarithm.
(2) Government green subsidy (GGS). The government's environmental protection, sustainable development, and low-carbon economy subsidies are called government green subsidies. In existing studies, green subsidies are mostly used to explain a company's investment in improving environmental pollution, and the cost of "green expenditure" and "pollution control expenditure" in the CSR report is used to measure the green investment of enterprises (Huang and Lei 2021). Since this study investigates the mechanism of green investment at the city level, we use government environmental expenditure as the basis for government green subsidies. Government environmental expenditures include water pollution control, air pollution control, soil pollution prevention and control, and investment in monitoring equipment. Next, we measure the level of government green subsidies by using the proportion of government environmental expenditures in total fiscal expenditures.

Control variable
Due to differences in the individual characteristics of different cities, to reduce the impact of other dimensional characteristics on the regression results, we select the following control variables to assist the regression. (1) Trade openness (OPEN). Higher trade openness is conducive to the entry of high-tech industries, the development of urban green technology industries, and the promotion of environmental pollution management . (2) Government intervention (GOV). The government conducts macrocontrol and capital guidance on the market through policy reform and improvement. It intervenes to a considerable extent in the green transformation of cities, which is conducive to reducing environmental pollution. (3) Educational basics (EDU). The higher the level of urban education is, the more conducive it is to generating and disseminating residents' environmental protection awareness (Qiu et al. 2021). (4) Economic development (GDP). A higher per capita GDP usually means a relatively larger economic volume and more production transactions, which is conducive to developing digital finance (Brink et al. 2016). (5) Urbanization level (URBAN). A higher level of urbanization can increase key production capacity through efficient green technologies while curbing environmental pollution (Müller et al. 2020). Table 2 summarizes the above variables and shows the specific calculation process of the control variables.

Model construction
To test the direct impact of digital finance on environmental pollution management, a fixed-effects model is selected for testing. At the same time, we controlled the time and city levels to reduce the bias caused by endogenous factors in the test results. The specific model is as follows: where Pol it represents the environmental pollution management level of city i in year t, DIFI it represents the digital financial level of city i in year t, i refers to the city, t represents the year, Control represents the control variable, θ represents the fixed effect of the city, δ represents the year fixed effects, and ε represents the random error term.
In addition, to test the intermediary role of green technology innovation and government green investment, this study draws on the research method of Wen and Ye (2014). It uses the mediation effect model to test the transmission mechanism. To ensure the accuracy of the test, we selected the Sobel test for further analysis. Specific models are as follows: Among them, M represents the intermediary variables, green technology innovation, and government green investment, and the rest of the variables are the same as in Model (6). If coefficients γ 1 , λ 1 , and λ 2 all pass the significance test of at least 5%, it means that the intermediary variable has a mechanism effect.

Measurement results of environmental pollution management
To visually display the spatial distribution of China's environmental pollution management level, this study uses ArcGIS software to perform spatial visualization processing on the urban environmental pollution management level in 2011 and 2021. Figures 2 and 3   have relatively high levels of environmental pollution control. This is perhaps due to two factors. On the one hand, there are differences in the level of economic development. Economic development in the central, southern, and northwestern regions may lag, and there are more opportunities for environmental protection and green industry transformation. With the adjustment of economic structure, these areas are more likely to adopt environmental protection technology and sustainable development models to reduce environmental pollution. On the other hand, resource endowments are different. The distribution of resources and industrial structure in different regions are quite different. The northwest region is rich in resources, which can develop more clean energy and reduce dependence on high-polluting industries. The southern region has manufacturing and foreign trade advantages and can introduce more environmental protection technologies, which is conducive to improving environmental pollution control. In addition, we found that the level of environmental pollution control in the Yangtze River Basin is higher than that in the Yellow River Basin. The reason is perhaps that the Yangtze River Basin is flat and rich in water resources. These geographical advantages make transmitting and diffusing pollutants in water bodies relatively easy and help control pollution. At the same time, the Yangtze River Basin has a large population, and people are more likely to feel the impact of environmental pollution. The public is more likely to actively participate in environmental protection activities and push the government to adopt more environmental protection measures. Although the level of environmental pollution control in the Yangtze River Basin is relatively high, there are still various environmental challenges and problems. Environmental pollution management is a long-term work that requires the joint efforts of the government, enterprises, society, and the public. Table 3 reports the descriptive statistics of the key variables of this study. From the perspective of the environmental pollution management level, the maximum value is 0.9984, the minimum value is 0.1088, and the maximum value is significantly greater than the average value of 0.7502, indicating that there are significant differences in the level of environmental pollution management among cities and a serious imbalance. At the same time, there are large differences in the dispersion of data in various dimensions of digital finance, indicating significant differences in the development level of digital finance among cities. The remaining control and mediating variables have different degrees of variation, but no obvious extreme values are found, and the values are all within a reasonable range. Table 4 reports the baseline regression results of digital finance on the level of environmental pollution management. Specifically, in column (1) of Table 4, the estimated coefficient of digital finance is significantly positive at the 1% level, indicating that digital finance significantly impacts the environmental pollution control index. With the development of digital finance, the level of environmental pollution management has been improved, and Hypothesis 1 has been verified. Digital finance can integrate many environmentrelated data, such as weather data, air quality data, and energy consumption data, and use data analysis and artificial intelligence technology to predict environmental change trends and risks. This will help detect environmental problems early, formulate corresponding environmental protection measures, and improve the management level of urban environmental pollution. According to columns (2) and (4)  of Table 4, we found that both the degree of digitization and the coverage of digital finance passed the significance test. The regression coefficients of digital finance are significantly positive at 5% and 1%. This shows that the digitalization degree and coverage of digital finance can promote improving urban environmental pollution management. The increase in the degree of digitization helps strengthen the diversification and convenience of digital financial services. It can track the carbon emissions of enterprises and individuals, establish a carbon emission trading platform, and reduce carbon emissions through carbon trading incentives. In addition, with the increase in service coverage, digital finance can help more cities carry out digital construction and realize intelligent management of urban energy, transportation, water resources, etc. These measures can effectively improve the efficiency of resource utilization and promote the control of environmental pollution. However, column (3) of Table 4 shows that the estimated coefficient of digital finance use depth on the environmental pollution management level is insignificant, indicating that the expansion of digital finance use depth cannot effectively promote the improvement of the environmental pollution management level. We analyzed the reasons for this and found that the analysis and forecast of digital finance depend on high-quality data. However, with the extension of the depth of digital technology, microenvironmental data are prone to problems of inaccuracy and incompleteness, and thus the application of digital finance will lose its accuracy. This may lead to incorrect decision-making and actions in environmental pollution management, which is not conducive to the improvement of the pollution management level.

Endogeneity problem
In existing studies, we found that digital finance has an endogenous problem of investment orientation. That is, digital finance can provide financing for environmentally friendly projects but may also provide funds for environmentally unfriendly projects. For example, investors may obtain high returns in industries that pollute the environment and thus tend to support these industries, leading to the aggravation of environmental pollution problems. To overcome this endogeneity problem, we use instrumental variable analysis. Specifically, instrumental variables are used to separate causality in the two-stage least squares method, and we choose Internet coverage (Inter) as an instrumental variable. The reason is that Internet coverage has both the characteristics of independence and correlation of instrumental variables. First, Internet coverage usually does not directly affect environmental pollution management, so it meets the independence requirements of instrumental variables. Second, although Internet coverage does not directly affect environmental pollution management, it may be related to the development level of digital finance. In some cases, areas with high Internet coverage may have easier access to digital financial services, thus affecting the role of digital finance in environmental governance. Regarding the calculation of Internet coverage, this study draws on the measurement method of Xie et al. (2018) and uses the proportion of total Internet port access in the city to the permanent population at the end of the year to measure it. Table 5 reports the results of the endogeneity tests. The instrumental variable Internet coverage (Inter) passed the weak instrumental variable test with an F value of 311.994. According to the regression results of the second stage, the coefficient of digital finance on the environmental pollution management level is significant at the 5% level, indicating no endogenous relationship between digital finance and the environmental pollution management level.

Robustness testing
To ensure the baseline regression results' robustness, we use three methods: replacing the dependent variable and data samples and replacing the model to conduct robustness tests.

Replacing the dependent variable
To avoid the error of the measurement method of the environmental pollution management level of the dependent variable from biasing the regression results, this study uses urban carbon emissions (CE) to replace the original dependent variable. Carbon emissions are selected as the replacement dependent variable because carbon dioxide is one of the main greenhouse gases and is closely related to climate change. In addition, carbon emissions have become an environmental indicator that is widely used in the worldwide, and monitoring carbon emissions can encourage companies to adopt energy-saving and emission-reduction measures and improve environmental pollution management (Lei et al. 2023). Similarly, we use Model (6) for regression testing. Column (1) of Table 6 reports the results of the robustness tests after replacing the dependent variable. The coefficient of digital finance is significantly negative at the 5% level, indicating that digital finance can significantly reduce carbon emissions in cities. In other words, digital finance can promote environmental pollution management, which is consistent with the baseline regression results.

Replacement sample
Furthermore, to eliminate the resulting error caused by sample selection bias, we used the method of replacing the sample to conduct a robustness test again. We eliminated four municipalities directly under the Central Government in China: Beijing, Tianjin, Shanghai, and Chongqing. Since the municipalities directly under the Central Government are important nodes in developing national strategies, their administrative management methods and economic development models are different from those of other cities. Considering the influence of exogenous factors, we test all samples after excluding the municipalities directly under the central government. Column (2) of Table 6 reports the test results. The coefficient of digital finance is significant at the 5% level. This shows that after excluding municipalities directly under the central government, digital finance still positively impacts improving environmental pollution management, which is consistent with the previous conclusion.

Replacement model
In the benchmark regression, this study chooses the fixedeffects model to test the impact of digital finance. In the fixedeffects model, we introduce city individual fixation. However, missing variables may still be missing, and the model does not include important factors affecting the dependent variable. If these omitted variables are related to the explanatory variables, the estimates from the fixed-effects model may be biased. To improve the robustness of the conclusions in this paper, this paper adopts the mixed effect model (OLS) and the random effect model (Re) to replace the original Model (6) for robustness testing. Random fixed-effects models allow individual effects to vary between random and fixed, thereby better coping with individual heterogeneity and temporal variation. This allows the model to respond flexibly to changes in time while capturing interindividual differences. Columns (3) and (4) of Table 6 report the test results after replacing the regression model. In the mixed and random effect models, the coefficients of digital finance are all significant at the 1% level, which is consistent with the benchmark regression model, hence obtaining the same result. This further strengthens the convincing power of the above conclusions.

Mediation effect analysis
This research explores the influence channels of environmental pollution management from the two perspectives of  green technology innovation and government green subsidies. Model (7) and Model (8) were used to analyze the influence channel, and Table 7 reports the regression results. The results of columns (1) and (2) show that the coefficients of digital finance and green technology innovation pass the significance test, indicating that digital finance can promote the improvement of green technology innovation, which in turn is beneficial to the development of environmental pollution management. The results of this test confirmed Hypothesis 2. This shows that digital finance supports the financing of sustainable development projects. Through green bonds, sustainable investment funds, etc., investors can invest funds in green technology innovation and control environmental pollution at the source. Columns (3) and (4) show that the coefficients of digital finance and government green subsidies are significantly positive at the 1% level. This shows that the government's green subsidies can play a mechanism role and that digital finance has a vital impact channel with respect to environmental pollution management. This result verifies Hypothesis 3. The reason is perhaps that digital finance can enhance the transparency and traceability of government green subsidies. Through technologies such as blockchain, the flow of subsidy funds can be tracked to ensure that it is accurately used for environmental protection projects, and thus it can improve the government's environmental pollution management level.

Heterogeneity analysis
Due to China's vast territory, there are large differences in climate and natural resources in different regions. At the same time, due to different macroeconomic policies, the economic development environment presents a differentiation trend. To explore whether there is heterogeneity in the impact of digital finance on the level of environmental pollution control in different locations, we divide the full sample into samples in the eastern region and samples in the central and western regions. Model (6) is used for regression. Columns (1) and (2) of Table 8 report the results of the positional heterogeneity analysis. It can be seen from the table that the digital finance coefficient in the central and western regions passed the 1% significance test, while the eastern region failed. This shows that the role of digital finance in promoting environmental pollution control is better in the central and western regions than in the eastern region. Analyzing the reasons, we found that improving environmental pollution management in the eastern region is relatively difficult due to factors such as high population density and rapid infrastructure development. In the central and western regions, green industry based on clean energy and green tourism has developed rapidly, and a relatively complete upstream and downstream green industry ecosystem has been formed, which will help improve the level of environmental pollution control. Next, we consider that minerals, energy, and other resources usually support resource-based cities. In economic development, these cities may face the challenge of environmental pollution because the extraction and processing of resources may lead to a large amount of waste and pollutant discharge. Therefore, is there any difference in improving environmental pollution management in resource-and non-resource-based cities? To test this conjecture, we refer to the research of Hou et al. (2019); resource-based cities refer to cities whose main industries are resource extraction (such as mining and forestry). We define this type of city as a resource-based city and mark it as "1" and the rest of the cities as "0." Then, we use Model (6) for sub-sample regression. Columns (3) and (4) of Table 8 report the regression results on city-type heterogeneity. Among them, the digital finance coefficient of resource-based cities is significant at the 1% level, while that of non-resource-based cities is insignificant. This shows that the promotion effect of digital finance on environmental pollution control in resource-based cities is better than that in non-resource-based cities. While verifying the above conjecture, it shows that digital finance helps promote the construction of smart cities. It can optimize urban layout, traffic flow, architectural design, etc. in resource-based cities to reduce environmental pollution, improve resource utilization efficiency, and make it more liveable and sustainable.

Conclusion
Digital finance is an important driving force for high-quality economic development, and it is also the core goal of green growth and environmental pollution control. Based on panel data of 287 Chinese cities from 2011 to 2021, this paper studies the impact of digital finance on environmental pollution management. The following conclusions were drawn: 1. Judging from the level of urban environmental pollution management, the country's overall level is relatively high, and the achievements in environmental pollution control have been remarkable in recent years. Specifically, the level of environmental pollution management in the northwest, central, and southern regions is relatively high, while that in the North Central region is relatively low. This shows that the overall difference in environmental pollution management is relatively large and requires coordinated development across the country. In addition, cities in the Yellow River Basin have a lower level of environmental pollution management than cities in the Yangtze River Basin. Therefore, the country should pay attention to measures to solve the inherent problems caused by geographical factors and, such as the sediment problem in the Yellow River Basin, to realize the coordinated development of regional environmental pollution management. 2. Digital finance can significantly improve the level of urban environmental pollution control, and both the degree of digitization and the breadth of coverage of digital finance can promote environmental pollution management. The intermediary effect test finds that, on the one hand, digital finance can promote urban green technology innovation, restrain environmental pollution at the source of production, and then promote urban environmental pollution management. On the other hand, digital finance helps the government to efficiently use green subsidies and promote urban environmental pollution through macrocontrol. 3. From the perspective of heterogeneity, the promotion effect of digital finance on environmental pollution management is significantly greater in the central and western cities than in the eastern cities. By classifying cities as a resource and non-resource cities, it is found that digital finance is more conducive to promoting resource cities. Resource cities usually rely on traditional industries such as resource extraction and processing, which in turn causes environmental pollution problems. Digital finance can support the establishment of an emission rights trading market and encourage companies to reduce pollution emissions through carbon emission quota trading and other mechanisms.

Policy recommendations
Based on the above research results, we put forwards the following policy recommendations to improve the current situation of environmental pollution management.
First, the above research conclusions conclude that digital finance can improve environmental pollution management. In this regard, the government can invest in developing digital infrastructure to support the application and promotion of digital finance, including high-speed Internet, mobile payment infrastructure, and data centers. In addition, the government should reasonably guide the flow of funds and encourage banks and financial institutions to provide low-interest loans and other financial support to support the development of green environmental protection projects. This helps speed up approvals and capital flow and promotes the rapid implementation of environmental protection investments. At the same time, financial institutions can use digital finance to develop environmental risk assessment tools to help the government identify possible environmental risks and impacts and take preventive and control measures to improve environmental pollution management.
Second, this research finds that green technology innovation and government green subsidies are important channels of influence for digital finance promotion. Therefore, the government should provide incentives for green technology innovation by formulating policies, reducing innovation risks, and encouraging companies and research institutions to invest in green field research and development. At the same time, the government should actively cooperate with other countries and international organizations to share experiences, cooperate in R&D, and accelerate the international dissemination and application of green technologies. In terms of green subsidies, the government can optimize the structure of green subsidies to ensure that funds are allocated reasonably. Invest more resources in areas with high environmental benefits and innovation potential. The government should ensure the transparency of the use and distribution process of green subsidies and establish a monitoring mechanism to prevent the abuse or waste of funds.
Third, by conducting heterogeneity analysis, we found that the promotion effect of digital finance in the western and central regions is poor, which may be caused by the education level and residents' financial literacy. In this regard, the government should organize financial knowledge training and promote financial education courses and other activities to improve residents' financial literacy in the central and western regions. This will help them better understand and use digital financial tools. Financial institutions should develop innovative financial products and services that meet the needs of the central and western regions and better meet the financial needs of the western and major regions. In addition, in the heterogeneity analysis, we found that resource cities are more conducive to improving environmental pollution management levels. Therefore, the government should establish a discharge permit and examination and approval system to permit and approve pollution sources to ensure that they meet the requirements of environmental regulations. At the same time, the environmental protection department should formulate and strictly implement pollutant discharge standards to ensure that the company does not exceed the prescribed discharge limit during the production process thereby reducing the release of pollutants.
The following limitations still exist in this paper. From a regional point of view, urban environmental pollution may have a certain spatial effect. This paper does not use a spatial econometric model that can examine spatial effects. In addition, the heterogeneity of digital finance may lead to differences in the effects, which we have not yet analyzed.
Addendum Due to space limitations, we can only show the results of environmental pollution management calculations for 31 provincial capital cities in the "Appendix" section. The environmental pollution management data of the remaining cities can be obtained from our author.
In future research, we can use spatial econometric models to test the spatial autocorrelation and spillover effects of environmental pollution management from the spatial perspective. Further empirical testing of other heterogeneity of digital finance is needed.