Abstract
The River Chief System (RCS) in China plays a crucial role in addressing transboundary water pollution (TWP), which is vital for achieving the Sustainable Development Goal of "clean water." This study examines the static and dynamic effects of the RCS on TWP using a difference-in-difference-in-differences (DDD) model and manually collected RCS data from 104 counties between 2007 and 2020. The results show that the RCS significantly reduces chemical oxygen demand (COD) by 15.9% and ammonia nitrogen (NH3-N) by 22.9% in border counties, though it has no notable impact on dissolved oxygen (DO). Additionally, RCS intensity contributes to a 13.8% reduction in COD. However, no sustained dynamic effects are observed for COD or NH3-N reduction over time. Several robustness checks confirm the validity of these findings. Heterogeneity analysis indicates that the RCS’s impact is more pronounced in counties with higher economic development, upstream locations, lower initial pollution levels, and younger local government leaders. This study not only provides critical insights for water resource management in China but also offers international relevance for addressing TWP governance challenges in transboundary water bodies.
Similar content being viewed by others
Introduction
Water pollution is worsening in various regions worldwide, with elevated pollution risks observed in Europe, India, China, South America, and certain areas of Africa1,2. According to the 2015 United Nations World Water Development Report, an alarming 90% of untreated sewage in developing nations is directly discharged into natural water sources3. The circumstances of China, which is the most populous developing nation and one of the world’s fastest-growing economies, serve as a significant representation of the global landscape, notably concerning issues such as water pollution4,5. Reports indicate that China faces a serious water pollution problem, with 28% of surface waters and more than 60% of groundwater classified as class IV or class V, rendering them unsuitable for human use6. At the same time, approximately 66% of China’s major coastal bays, encompassing a total area of 57,000 km2, fall into the class IV or class V category7. Primarily, it has been reported that China has experienced the loss of 60,000 lives due to illnesses caused by contaminated water8. Therefore, the urgency of water pollution control is evident because it is characterized by its nonexclusive and non-competitive nature9,10.
According to data from the Ministry of Ecological Environment in China, in 2019, among the 32 provincial boundary sections, 22.8% exhibited class IV–V water quality, and 4% were categorized as inferior class V11. Managing transboundary water pollution (TWP) can be intricate because multiple jurisdictions or countries are involved, potentially resulting in regional conflicts12. The establishment of the river chief system (RCS) in China represents a form of institutional exploration and innovation undertaken by local authorities to solve water environmental problems13. In 2007, the blue algae outbreak in China’s Tai Lake in Wuxi, Jiangsu Province, prompted the introduction of the RCS in Wuxi14. In 2016, the central government issued a nationwide directive for the implementation of the RCS in all prefectures by the end of 2018 due to a substantial improvement in the water quality of Tai Lake15. The mechanism designates the heads of the Party and government at every level as "river chiefs," responsible for water environmental governance in their domains16. River chiefs face dismissal from their leadership positions if major water pollution-related incidents occur in the jurisdictions that they govern17. However, the river management system also incurs systemic costs in practice, such as inefficiencies in governance arising from overlapping administrative functions18, and heightened local financial burdens due to a reliance on a single resource stream19 and the lack of market-based mechanisms20. Therefore, its overall effectiveness remains a subject of debate.
The full-scale adoption of the RCS has led to a significant increase in the amount of academic research focused on assessing its effectiveness. However, this surge in studies has yielded conflicting and contradictory findings. Prominent research demonstrates the effectiveness of the RCS in water environment improvement21,22. In contrast, alternative research has suggested that the RCS tends to have short-term influence and functions as an immediate tool14,19. While quantitative assessments of the effectiveness of the RCS in tackling TWP in China are lacking, research examining the impact of the RCS on TWP and the heterogeneity in this impact has crucial policy implications. First, considering the severity of TWP and its significant role in water environmental governance11, addressing the TWP issue can aid China in attaining the global Sustainable Development Goals set in the 2030 Agenda. Second, water pollution issues in bordering bodies of water are often influenced by different administrative jurisdictions, government agencies, and local interests, making the management of these issues highly challenging23,24. This underscores the significance of studying the impact of the RCS on the management of TWP, as this topic represents a pivotal issue in China’s water resource management and holds international relevance25.
Therefore, utilizing manually collected RCS data from 104 transboundary counties across China spanning from 2007 to 2020, we employ a systematic analytical framework with the difference-in-differences-in-differences (DDD) estimator to assess the impacts of the RCS on TWP and their underlying heterogeneities. The empirical findings affirm that the RCS policy effectively addresses TWP, and a battery of robustness tests corroborate these results. Additional research suggests that the economic development level, geographic location, initial levels of water pollution and county leader age are all important factors in determining the success of RCS policies. This study offers policy recommendations to effectively navigate the conflict between sustained economic growth and TWP through the development of scientifically sound and reasonable environmental policies.
The novelty of this research is reflected in three main aspects. Firstly, it is noteworthy that existing research has extensively explored the effectiveness of the RCS in improving water quality and protecting the water environment in China20,26. However, to the best of the authors’ knowledge, no prior study comprehensively investigates both the static and dynamic effects of the RCS on TWP. This gap highlights the need for a deeper understanding of the RCS’s short-term and long-term impacts on TWP. Secondly, this paper introduces an innovative approach to measuring TWP by analysing diurnal variations in pollution at boundary monitoring stations. Unlike traditional methods that rely on static or aggregated data, this approach captures the dynamic nature of water quality fluctuations, offering a more accurate representation of TWP patterns. By using high-frequency, time-specific data, it enhances monitoring precision, informs targeted interventions, and helps resole cross-boundary disputes. Thirdly, building upon the estimation of RCS effect, we analyse the factors influencing its effectiveness in mitigating TWP, including economic disparities, geographical positioning, initial pollution levels, and the age of local leaders, aiming to provide tailored insights for optimizing the RCS’s design and application across diverse contexts.
Background
Water pollution in China
The United Nations’ Sustainable Development Goals (SDGs) underscore the pivotal role of ensuring sufficient water quality as a fundamental component in attaining sustainable water and sanitation27. Surface water pollution is particularly problematic in China, recent endeavours to enhance water quality through heightened investment in water pollution abatement notwithstanding. The China National Environmental Monitoring Centre (CNEMC) provides data from the National Automatic Monitoring System. According to these data, in 2020, 13.6% of the country’s surface water was categorized as class IV, 2.4% as class V, and 0.6% as inferior class V. Consequently, a combined 16.6% of surface water was deemed unsuitable for drinking.
Figure 1 illustrates key water pollution indicators in China, namely chemical oxygen demand (COD), dissolved oxygen (DO), and ammonia nitrogen (NH3-N), presenting anomalies and 5-year moving averages. Daily water pollution data on these three indicators were used to compute the annual data, which were subsequently normalized to the mean reference period. Low-frequency variability was computed through a moving average with a period of 5 years. As shown in Fig. 1, COD and NH3-N exhibit a declining trend, notably with a gradually significant decrease in recent years. In contrast, DO has shown an increasing trend in recent years. These trends indicate that the water quality in China has gradually improved steadily over the past few years.
Annual mean specific indicators (COD, DO, NH3-N) of water pollution anomaly of monitoring stations in China from 2007 to 2020.
The RCS in China
In response to the urban water supply crisis triggered by widespread cyanobacterial blooms in Lake Tai, the Wuxi municipal government initiated the RCS in August 2007, issuing an official document titled “Decisions on establishing the RCS and forging a comprehensive governance system managing lakes, reservoirs, and springs”. This document not only outlined accountability mechanisms and criteria for performance assessment but also explicitly mandated government and party leaders under the jurisdiction of Wuxi to serve as “river chiefs”. Drawing from nearly a decade of exploration and practical insights, the Central Committee of the Chinese Communist Party issued the official document "Opinions on Promoting the River Chief System" in December 2016. This document explicitly mandated the establishment of the RCS by the end of 2018.
As shown in Fig. 2, following an extensive period of pilot initiatives and active implementation, China has established a comprehensive four-tier structure (covering provinces, municipalities, counties, and townships) for the RCS. By June 2018, more than 300,000 river chiefs had been designated across various administrative levels, including the provincial, city, county, and township levels. At every tier, appointed river chiefs are collectively responsible for water environment improvement, which constitutes a pivotal factor in evaluating and promoting officials. The delineation of responsibilities, authorities, and incentives for the environmental stewardship of rivers within this four-tier RCS framework is pivotal. This framework serves as a compelling model, exhibiting tangible improvements in environmental management practices, thus acting as a guiding light for similar initiatives.
A comprehensive four-tier structure for the River Chief System (RCS).
Materials and methods
To depict the analysis of the effect of the RCS on TWP, a flow chart outlining the analysis methods is presented in Fig. 3. Initially, we gathered data on the RCS and TWP. Subsequently, utilizing these data, we examined both the average and dynamic effects of the RCS and the implementation intensity of RCS on TWP. Finally, we conducted an analysis to explore the heterogeneity within the dataset.
Research flow chart of analysis method.
Empirical model
The DID method, employed to juxtapose a treatment group against a control group, serves to ameliorate deviations. In our effort to evaluate the impact of the RCS on the reduction in TWP, we introduce the difference-in-difference-in-differences method (DDD), an extension of the DID method that necessitates the inclusion of an additional control group in the estimation. In contrast to traditional multiple linear regression, the DDD method establishes a more stringent causal identification framework that can address selection bias.
The average effect of the RCS on TWP
Due to the inevitable increase in pollution at boundary monitoring stations caused by upstream-to-downstream pollution discharge, this paper aims to use boundary stations with pollution as the experimental group and those without pollution as the control group. Subsequently, the study employs the DDD model for estimation.
where \({WaterPollution}_{ict}\) represents the indicators of water pollution (COD, NH3-N, DO) for station \(i\) in county \(c\) at year \(t\); \({RCS}_{c}\) is a binary indicator, taking the value of 1 if county \(c\) has implemented RCS and 0 otherwise; \({Border}_{i}\) is a location dummy that equals 1 if station \(i\) is on the border of a county and 0 otherwise; \({After}_{t}\) is a post-treatment binary indicator, taking the value of 1 if \(t>{t}_{i0}\), where \({t}_{i0}\) is the year in which county \(c\) implemented the RCS, and 0 otherwise; and \({X}_{ict}\) represents a set of county-level control variables. \({\mu }_{i}\) represents the station fixed effect, capturing the time-invariant distinctions among stations; \({\vartheta }_{t}\) denotes the year fixed effect; and \({\varepsilon }_{ict}\) represents the error term. The focal coefficient is \({\alpha }_{1}\), which represents the average effect of the RCS on TWP.
The average effect of RCS intensity on TWP
To further investigate whether the actual implementation intensity of the river chief system (RCSI) can contribute to alleviating TWP, we subsequently substitute the core explanatory variable \({RCS}_{c}\) with \({RCSI}_{c}\). The model is as follows:
where \({RCSI}_{c}\) is the county-level RCS intensity and \({\beta }_{1}\) is the coefficient of interest, which represents the impact of RCS intensity on TWP. All the other variables are consistent with Eq. (1).
The dynamic effects of the RCS on TWP
Subsequently, we examine how the RCS binary variable affects TWP over time. We modify Eq. (1) by incorporating a set of dummy variables to capture the annual impact of the RCS binary variable on TWP:
where \({RCS}_{st}^{k}\) represents a set of “event-time” dummy variables that take the value of 1 when RCS implementation occurs \(k\) years later in a county; here, \(st\) is the year when the county implemented the RCS, with \(k\ge 0\); and \({RCS}_{st}^{k}\) takes the value of one if the county is in the \(k\) year following RCS implementation.
The dynamic effects of RCSI on TWP
Equation (3) evaluates the dynamic effects of the RCS binary variable on TWP. Finally, to further scrutinize the dynamic effects of RCS intensity on TWP, we substitute \({RCS}_{st}^{k}\) with \({RCSI}_{st}^{k}\):
where \({RCSI}_{st}^{k}\) represents the RCS intensity \(k\) year after the county implemented the RCS. The other variables are consistent with Eq. (3).
Variables
Independent variable: RCS
RCS dummy: Implement the RCS or not
We manually compiled the implementation status of the RCS in the counties where 109 monitoring stations were located from 2007 to 2020. In the initial phase, we conducted searches on the official websites of each county government using keywords such as “RCS” or “river chief” to gather official documents related to the RCS, including local regulations and government provisions. Subsequently, through manual organization, we summarized information on whether each county had implemented the RCS and the specific year of implementation. Then, we conducted searches on the Baidu Encyclopaedia using keywords such as “RCS” or “river chief” to retrieve news reports, and we manually organized information on the implementation of the RCS in each county. Finally, we cross-referenced the two sets of RCS implementation data to ensure the accuracy of the manual sorting results.
The final compilation of fundamental information on the implementation of the RCS is depicted in Fig. 4. As shown, in 2013, the system was adopted by only 5 sample counties, but by 2017, this number had risen to 100, with all counties adopting the system in 2018.
The number and percentage of counties implementing River Chief System (RCS) from 2007–2020.
RCS intensity: The implementation intensity of the RCS
For the variable RCS intensity, the greater the number of sentences in the government work report mentioning the RCS is, the more it reflects the government’s emphasis on the system28. We systematically identified RCS-related keywords in the annual government work reports of each county, encompassing terms such as “river chief system”, “river chief”, and “lake chief”. Subsequently, we located the sentences containing these keywords and quantified the word count within these identified sentences. Finally, we computed the ratio of the words related to the RCS to the overall word count in the government work report and considered this ratio to be indicative of the implementation intensity of the RCS.
Dependent variable: Transboundary water pollution
The National Automatic Monitoring System of Surface Water Quality (NAMS) is an initiative that was launched by the Chinese National Environmental Monitoring Centre (CNEMC) in 2004. It collects weekly data on four water pollution indicators, water quality grade, dissolved oxygen (DO), chemical oxygen demand (COD), and ammonia nitrogen (NH3-N), for the major river basins in China. Among these four indicators, the water quality grade is an overall assessment of water pollution that includes too much information and thus may result in biased estimation results. As a result, we chose yearly data on the three-remaining metrics—DO, COD, and NH3-N—as proxies for water pollution from 2007 to 2020. It is crucial to emphasize that elevated values of COD and NH3-N indicate more polluted water, whereas lower DO values signify higher degrees of water pollution.
Boundary variables
Previous studies on water pollution in China have adopted 5 kms as the spatial scale for county-level analysis17,29, aligning with the governance characteristics of counties. Furthermore, surface water pollution exhibits spatial dissipation, with emissions having diminishing impacts as distance increases. Based on these considerations, we categorized the monitoring stations into two distinct groups: transboundary and interior monitoring stations. We calculated the spatial proximity of each monitoring station along the river to the county boundary by utilizing the geographic coordinates of these stations using QGIS version 4.30. As shown in Fig. 5, monitoring stations located within a 5-km range are designated trans-border stations, while those exceeding the 5-km threshold are categorized as interior stations. Then, we matched each monitoring station to the county, leveraging precise geographic information, which enabled us to align the data with the corresponding county. Subsequently, we integrated the data from monitoring stations and county-level data spanning from 2007 to 2020. Table 1 clearly shows that for the majority of years, the concentrations of COD and NH3-N at transboundary monitoring points exceeded those at interior monitoring points, while the opposite trend was observed for DO, signifying the existence of a distinct TWP phenomenon in China.
Formation process of transboundary water pollution.
To gain a preliminary understanding of the distributions of the three water pollution indicators, we plotted kernel density graphs before and after the implementation of the RCS. As depicted in Fig. 6, the kernel density disparity of DO before and after RCS implementation is prominently significant, displaying a distinct rightward skew. Conversely, the post-RCS indicators for COD and NH3-N manifest a noticeable leftward skew when contrasted with their respective pre-RCS counterparts. While potential uncontrolled confounding factors should be considered, this observed trend strongly indicates that the RCS significantly alleviates TWP.
Kernel Density of water pollution indicators (COD, NH3-N and DO) in boundary stations before and after the implementation of River Chief System (RCS).
Control variables
Considering the complex and diverse factors influencing water pollution, to mitigate omitted variable bias and enhance identification accuracy, we controlled for other variables affecting water pollution based on relevant research30. The principal control variables selected encompass GDP per capita, population density, the county land area, night-time luminosity, the presence of national or provincial scenic spots and nature reserves within the county, the average annual temperature, and annual rainfall. In the robustness check, we incorporated controls for specific policy details, such as the Ecological Civilization Demonstration Area (ECDA) and Livestock Environmental Regulations (LERs). Furthermore, in recognition of the heterogeneity in political incentives among various county leaders, we systematically complied detailed biographical information on the age of county leaders from the China Statistical Yearbook (County-level), Baidu Encyclopaedia, and the China Vitae internet service. The specific definitions of the variables mentioned above are summarized in Table 2.
Data and sample
Sample selection
Our research samples are derived from 104 counties that are home to 109 national automatic monitoring stations, which serve as the focal points for our research. These stations are part of the NAMS, established by the Ministry of Environmental Protection (MEP) in the 1990s, whose primary purpose is to gather information on surface water quality by deploying 148 national automatic monitoring stations across the country. For the following reasons, we excluded 39 stations out of the 148 stations from our sample. Initially, we excluded 28 stations that are not situated in rivers, leaving 120 stations located specifically in rivers. These 28 stations are located in lakes or reservoirs, which exhibit hydrological distinctions; moreover, 11 monitoring stations with substantial data gaps were eliminated. Subsequently, we matched the station data to each county and amalgamated them with other variable data, yielding data for 104 counties based on comprehensive geographic information for 109 monitoring stations. The geographical locations of the sample stations are depicted in Fig. 7.
The Spatial Distribution of Surface Water Quality Monitoring Stations in China (created by QGIS 4.30 https://qgis.org/). Note on copyright: The base map in Fig. 7 of this paper is derived from the Standard Map Service System administered by the Ministry of Natural Resources of the People’s Republic of China (http://bzdt.ch.mnr.gov.cn/). These standard maps were compiled in strict compliance with internationally recognized cartographic standards for national boundaries delineation between China and other sovereign states. The system explicitly authorizes the utilization of such cartographic resources for multiple purposes including but not limited to: news publicity materials, illustrations in publications and periodicals, background visuals for advertising displays, base maps for handicraft designs, as well as serving as reference base maps for compiling publicly accessible maps. The cartographic foundation of Fig. 7 is exclusively sourced from this certified system.
Data sources
Before commencing our analysis, we conducted data pre-processing through the following steps. Initially, the NAMS of the CNEMC provided weekly automatic water quality monitoring reports up to 2018 and released real-time data at four-hour intervals daily. Based on these data, we calculated the annual average for the three water pollution indicators at each sample station. Subsequently, leveraging the comprehensive geographic data associated with each monitoring station, we were able to accurately pinpoint the corresponding county in which each station was located; then, with the geographical location of each station, we could align and consolidate the data within each respective county. The county-level GDP per capita, population density, and land area data were gathered from the China Statistical Yearbook (county-level) and the China City Statistical Yearbook. Night-time luminosity data came from the Harvard Dataverse. Data on nature reserves and scenic spots were acquired from the Ministry of Ecology and Environment of the People’s Republic of China (MEPC). Temperature and precipitation data were obtained from the China Meteorological Data Sharing Service System.
Results and discussion
Average effect of the RCS on TWP
Effects of the RCS dummy on TWP
Table 3 presents the principal findings of this study obtained by analysing the impact of the RCS on TWP using the DDD model. Columns (1) and (3) of Table 3 reveal a positively signed and statistically significant coefficient for the DDD method at the 5% level. This finding suggests that after accounting for other influencing factors, the RCS policy results in 15.9% and 22.9% reductions in COD and NH3-N, respectively, in the boundary counties. Furthermore, our findings align with those reported in the literature31. reported a significant reduction of 3.7% in firm-level COD emissions attributable to the RCS32. stated that RCS implementation correlated with reduction of 0.927 mg/L and 0.069 mg/L in COD and NH3-N, respectively. However, Column (2) of Table 3 presents the regression results of the impact of the RCS on the DO indicators, revealing no significant impacts of the RCS on DO in the border areas. One plausible explanation is that the RCS primarily restricts the discharge of industrial, agricultural, and domestic sewage, while DO typically serves as a survival indicator for aquatic organisms7,33. Differently19, discovered that after implementing the RCS, there was an increase of 0.369 units in the average DO. They explained that during RCS implementation, local governments may prioritize surface-level actions, such as removing algae and litter, over addressing the root causes of water pollution, facing criticism for being cosmetic and neglecting core issues related to polluted and malodorous water.
Effects of RCS intensity on TWP
We further extend our analysis by gauging the intensity of RCS implementation with a continuous variable. Table 4 shows that RCS intensity continues to exert a statistically significant effect on COD reduction within boundary areas. Specifically, a 1-unit increase in RCS intensity leads to a corresponding 13.8% reduction in COD. Notably, the mean RCS intensity in our research sample is only 0.58%. Consequently, a 1-unit increase in RCS intensity represents a substantial change. This finding suggests that the greater the implementation intensity of the RCS is, the better the treatment effect on water pollution. However, the coefficients for DO and NH3-N are not statistically significant. This lack of significance may be attributed to local economic activities, land-use patterns, or specific industrial processes, which could introduce variations in the effectiveness of RCS intensity across diverse water pollution indicators17.
Analysis of the dynamic impacts of the RCS on TWP
To provide a more precise evaluation of the impacts of the RCS, we continue to explore the dynamic impacts of the RCS, as illustrated in Table 5. Post-RCS implementation, there was an immediate substantial decrease in COD, observed for only four years, with the statistically significant reduction effect ceasing after the fourth year and beyond. There was a nuanced difference in that following RCS implementation, there was a notable reduction in NH3-N in the first year. Notably, in the fourth year of RCS implementation and beyond, the reduction in NH3-N no longer maintained statistical significance. The RCS dummy still does not have a significant impact on DO over most years, aligning with the results of the average effects discussed previously. Table 6 delineates the outcomes of the dynamic impacts of RCS intensity, and the trends closely resemble those of the RCS dummy, further confirming our conclusions.
This nonlinear dynamic change indicates that the RCS has no dynamic effect on TWP, similar to the findings reported by previous researchers21,34. A plausible economic explanation for this phenomenon is the tendency of local Chinese governments to prioritize short-term pollution control measures, such as temporarily shutting down polluting factories or removing waste from waterways. While these actions may yield immediate improvements in water quality, they often focus on superficial results rather than addressing root causes19. The emphasis on rapid compliance under high-pressure enforcement risks undermining long-term sustainability, as deeper, systemic solutions are often overlooked in favor of quick fixes18. However, despite witnessing some initial successes, local authorities continue to prioritize economic growth, which poses a persistent obstacle to the effective and sustained implementation of the RCS35.
Robustness checks
The credibility of the aforementioned results is confirmed through thorough robustness tests examining the following aspects:
Parallel trend test
Event study
A fundamental hypothesis for DDD model estimation is the presence of a common trend: the patterns of changes in water pollution indicators at boundary stations in RCS and non-RCS counties are assumed to be identical. Hence, in the initial diagnostic test within this section, an event study was conducted based on the approach outlined by36. Specifically, the adopted estimation formula is as follows:
\({RCS}_{st}^{n}\) is a set of dummy variables that equals 1 if the year is \(n\) year before/after the county implemented the RCS in year \(st\) and 0 otherwise; \(n=t-st\), where \(n=\{-5,-4, -3, -2, -1, 0, 1, 2, 3\}\). We categorize the lead policy dummies that are more than 5 years before the implementation of the RCS as \(n\le -5\) and those that are more than 3 years after the implementation of the RCS as \(n\ge 3.\) \({\alpha }_{1}\) captures the difference in water pollution indicators at the boundary stations between RCS counties and non-RCS counties. If the preceding policy dummies exhibit no statistically significant outcomes, this suggests a lack of evidence regarding differential trends before the introduction of the RCS.
Common trend testing
While the disparity in COD and NH3-N at boundary stations between RCS and non-RCS counties is significant in Sect. “Effects of the RCS dummy on TWP”, we need to verify that the pre-trend is identical for both RCS and non-RCS counties by performing a set of common pre-trend tests. As demonstrated in Fig. 8, the coefficients of all the indicators (COD, DO, and NH3-N) are statistically nonsignificant within a 90% confidence interval before the RCS was implemented, which indicates that RCS and non-RCS countries exhibit the same trend in the absence of the RCS, fulfilling the common pre-trend assumption.
Parallel trend test.
Placebo test
To eliminate the potential impact of unobserved variables on the assessment of RCS effects, we performed a placebo test using a methodology inspired by37. We utilized a random selection process for determining the implementation year of the RCS and identifying the monitoring sites influenced by the RCS. In this test, we performed 500 random selections in the stimulation and re-regressed the benchmark model; the distribution of 500 simulated DDD estimators is depicted in Fig. 9. The coefficient obtained based on the random sample estimation is distributed around 0, while the coefficient of the baseline regression estimation (-0.159 for COD, 0.005 for DO, and -0.229 for NH3-N) is completely independent of the coefficient distribution. These findings show that the estimation results are unlikely to be influenced by unobservable factors at the provincial and temporal levels, underscoring the reliability of the results.
Placebo test.
Controlling for water pollution that pre-exists the RCS
Local governments possess the autonomy to independently decide when and where to adopt the RCS, guided by their economic and environmental development and objectives. For example, if a county experiences more pronounced water deterioration issues, it may exhibit a greater propensity to implement the RCS. Hence, to account for the potential influence of pre-RCS water pollution levels, we incorporate pre-RCS water pollution measures as supplementary variables in the benchmark model. The analysis presented in Table 7 demonstrates that the RCS and its implementation intensity still have a significant influence on the reduction in COD and NH3-N concentrations, closely aligning with the outcomes of the baseline model. These results imply the absence of apparent selection errors in our baseline model, thereby allowing us to dismiss the possibility of a potential endogeneity issue.
Confounding factors
In addition to the RCS policy, various other environmental instruments and measures were in effect when the RCS was implemented22,38,39, potentially introducing distortions affecting the accuracy of the baseline regression results. The Livestock Environmental Regulations (LERs) and the ECDA policy are two policies that should be excluded because they might impact TWP mitigation. Initially, the government of China, which is the world’s leading producer of livestock, responded to the deterioration of water quality linked to livestock production by instituting LERs in 200640. Following41, our methodology introduced one binary variable into our baseline model to represent whether a county had implemented LERs, thus isolating the influence of this simultaneous policy. Second, the enactment of the ECDA policy in 2013 included a spectrum of regulations focused on water pollution control; as of 2020, this policy had been implemented in 270 counties nationwide. Hence, we introduced another binary variable into our baseline model to represent whether a county is situated in the ECDA, mitigating the impact of this policy.
The results from the analysis, considering the potentially confounding policies in Table 8, demonstrate that the effect of the RCS on COD and NH3-N reduction remains significant, even with the inclusion of the two other policy dummy variables. This finding implies that the RCS does indeed contribute to addressing TWP.
Excluding the effect of extreme values
One more potential concern about the benchmark estimation is that extreme values may influence the accuracy of the estimation results. For instance, within our sample, certain stations are located near sewage facilities, resulting in significantly greater water pollution than that at other stations. To address this issue, we removed 1% and 2.5% of the extreme values from the water pollution indicators to mitigate their impact.
Table 9 illustrates the influence of the RCS on TWP after the removal of extreme values. The significance of the RCS on COD and NH3-N persists, indicating that the impact of the RCS on TWP remains unaffected by extreme values. This finding reinforces the robustness of our conclusion.
PSM-DDD estimation
We utilized the propensity score matching (PSM) method, following42. PSM-DID estimation effectively addresses self-selection issues, aiming to minimize sample disparities by selecting a comparable non-RCS county for each RCS county. First, to overcome the bias caused by different initial conditions, we used the PSM method to match two groups of counties with the nearest neighbour in the callipers. When applying the PSM-DDD method, logit regression was employed for score computation, and the comprehensive definitions of these covariates are outlined in Table 10. The results of the balance test indicate notable differences in covariates between counties with boundary stations and those with interior stations before matching. However, we find no significant differences in covariates after matching, which ensures a high degree of similarity between the two groups of counties. When applying the PSM method, we utilized nearest-neighbour matching to assess the robustness of the impact of the RCS on TWP mitigation.
Table 11 reveals the effect of the RCS on TWP employing PSM-DDD identification. Notably, the coefficients of RCS remain statistically significant, suggesting that RCS can reduce COD and NH3-N.
Heterogeneous effects
In conducting heterogeneity analysis, our objective is to discern the types of counties most significantly influenced by the RCS, making it possible to customize more precise policies for distinct county groups. Furthermore, if the effect of the RCS on TWP varies across different sample subsets, our conclusions are reinforced.
Heterogeneous effect across economic development levels
Higher levels of economic development can enhance the impact of the RCS on TWP by correlating with increased investments in environmental pollution control and thereby improving the effectiveness of environmental regulations43. To assess the potential influence of economic development levels on the RCS effect, we categorize the sample into higher and lower economic development groups. The higher and lower economic levels of the samples are determined by the median GDP per capita.
According to the analysis, as shown in Table 12, both the RCS dummy and RCS intensity exhibit a significantly negative impact on COD in counties with higher economic development, while they do not have a significant influence on any indicator in counties with lower economic development. We propose that the varied results for different pollutants highlight their distinct sources. COD primarily originates from industrial sectors, dominating the local economy, while NH3-N originates from agricultural fertilizers and poultry farming. Thus, RCS implementation has a more significant impact on COD in economically developed regions21. This finding aligns with the findings of7. One probable explanation for these findings is that local governments in areas with higher and lower economic development may have divergent environmental protection objectives. In counties with lower economic development, where the primary focus is often on fostering economic growth to boost income, the practical implementation of environmental regulations such as the RCS is significantly curtailed44. Environmental Kuznets curve (EKC) theory states that the economic level has an inverted U-shaped relationship with the environment45. In contrast, counties with higher economic development have more funds and more advanced technologies, and they are more motivated to implement the RCS to protect the environment, resulting in better policy effects46.
Heterogeneous effect across geographic locations
The heterogeneity in geographical locations among counties in China poses challenges in promoting equitable environmental regulatory policies, such as the RCS47. We further subdivide the boundary stations into upstream and downstream counties. Within the boundary stations, those positioned in proximity to a river source are designated upstream samples, whereas those situated near an estuary are categorized as downstream samples. Among the 69 boundary stations, 32 are positioned upstream, and 37 are situated downstream.
According to the findings in Table 13, the RCS dummy leads to an average reduction of 18.9% and 26.6% in COD and NH3-N, respectively, in the upstream area samples, and RCS intensity reduces COD by 25.9%. However, the effects in downstream areas are not statistically significant. This observation is consistent with the findings of previous studies, as reported by28. This result can be explained as follows. First, pollution in upstream areas not only harms the local environment but also damages downstream regions. To ease the burden of accumulated pollution downstream, the Chinese government requires upstream regions to economically compensate for emissions to downstream subregions. Faced with these significant incentives, local governments in upstream areas are more motivated to tackle water pollution48. Second, the lower motivation for water pollution control in downstream regions compared to upstream regions stems from their predominant location along political borders. In these areas, the benefits of pollution control are partially absorbed by neighbouring administrative jurisdictions, thereby diminishing the overall net benefit49.
Heterogeneous effect across initial levels of water pollution
The impact of the RCS could be influenced by the initial pollution level. As suggested by50, areas characterized by elevated pollution concentrations tend to demonstrate a greater demand for environmental improvement. Other research also reveals that a pronounced initial concentration of water pollution results in an increase in the participation of farmers in the protection programme immediately upon the commencement of environmental policy implementation51. Hence, we utilize DO, COD, and NH3-N as indicators and split the sample into two groups based on the median values of these indicators. Counties with COD (or NH3-N) higher than the median or DO lower than the median are considered to have high initial pollution levels, while counties with COD (or NH3-N) lower than the median or DO higher than the median are considered to have low initial pollution levels.
The findings, presented in Table 14, indicate a significant reduction in COD of 10.8% due to the RCS and a reduction in COD of 12.3% due to RCS intensity in counties with initially low pollution levels. However, the RCS fails to significantly diminish the concentrations of the indicators in counties marked by high initial pollution levels. One possible reason is that areas with initially higher levels of water pollution may adopt pollution control measures that harmonize with economic growth because they may focus on increasing economic development29,52. In contrast, in areas with initially lower levels of water pollution, where there is already heightened environmental awareness53, the local government’s sensitivity to environmental pollution greatly increases the effectiveness of the RCS in these areas.
Heterogeneous effect across county leader age samples
Consistent with the study conducted by17, it has been demonstrated that a prefectural city headed by a leader with strong political motivations, often indicated by an age younger than 57 years, is inclined to implement regulations with greater rigor. Building on this logic, the age of 53 serves as a critical threshold for county-level leaders, as officials aged 53 or older in a given year are typically considered to have entered a “promotion stagnation period” and are thus categorized as “having weak political incentives”. In contrast, those younger than 53 are seen as “having strong political incentives,” as they are more likely to pursue innovative policies and reforms to advance their careers within the limited promotion window. This distinction aligns with the broader socioeconomic and political context, where younger leaders are increasingly valued for their adaptability to innovation-driven governance and longer expected tenures54. Consequently, we conduct a regression by dividing the boundary sample into two groups: Counties with leaders aged 53 or older and counties with leaders younger than 53.
Table 15 illustrates that in counties with a leader younger than 53 years, the RCS dummy results in a 17.9% reduction in COD and a 28% reduction in NH3-N, and RCS intensity results in a 24.9% reduction in COD; however, the coefficients fail to achieve statistical significance in counties led by individuals aged 53 or older. Since the proposal of the Scientific Outlook Development Concept by China in 2003, the importance of clean surface water has been repeatedly emphasized17. To achieve the specified goals for improving surface water quality, the central government has set pollution control requirements for each province. These environmental objectives have become crucial criteria for evaluating local politicians’ prospects for promotion. Younger leaders in townships enforce the RCS more rigorously due to their strong political promotion incentives, while older leaders who are ineligible for further promotion may not prioritize surface water quality improvement.
Conclusion and policy implications
Managing TWP is frequently intricate and demanding because multiple jurisdictions or countries are involved. Based on an unbalanced panel dataset comprising 1252 observations across 104 counties from 2007 to 2020, COD, DO, and NH3-N, which are indicators of water pollution, were employed to quantify TWP, and the DDD estimator was adopted to assess the impact of the RCS on TWP and the heterogeneity in this impact. The key findings of this research are outlined as follows: (1) The RCS dummy significantly contributes to addressing TWP, resulting in decreases of 15.9% and 22.9% in COD and NH3-N, respectively, with the intensity of RCS implementation being associated with a 13.8% decrease in COD, as validated through a series of robustness checks. (2) There is a nonlinear trend in the dynamic impacts of the RCS dummy on COD and NH3-N following the implementation of the RCS, lasting for four years, indicating the dynamic impact of the RCS. (3) Subsequent investigations reveal that the correlation between the RCS policy and TWP is associated with the regional economic development level, geographic location, the initial water pollution levels and the age of government leaders.
This study offers valuable policy recommendations for effectively reconciling the tension between achieving the SDGs and improving the water environment. Specific to the RCS policy, as mentioned earlier, the reason for the long-term poor effect of the RCS on COD and NH3-N may be the government’s band-aid remedy pollution control strategy. Therefore, scientific feedback mechanisms and measures should include conducting regular inspections, spot checks, and evaluations of water pollution control work in various regions; punishing leaders who damage the environment; and incorporating assessment indicators related to the actual effect on environmental improvements. Additionally, both the central government and county authorities should incentivize lower-level governments to establish collaborative governance institutions regionally. Information should be publicly accessible through government websites and other platforms to ensure that market players can stay informed about the cross-domain environmental situation in a timely manner. Finally, there is a need to enhance the management and maintenance of border monitoring stations, along with the installation of additional pollution monitoring stations at county boundaries. This measure aims to enhance the monitoring of TWP and to act as a deterrent against the illegal discharge of pollution by environmental violators.
Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
References
Rather, R. A. et al. Seasonal fluctuation of water quality and ecogenomic phylogeny of novel potential microbial pollution indicators of Veshaw River Kashmir-Western Himalaya. Environ. Pollut. 320, 121104 (2023).
Ferreira, D. M. et al. Modeling transport and fate of metals for risk assessment in the Parauapebas river. Environ. Impact. Assess. Rev. 102, 107209 (2023).
WWAP (United Nations World Water Assessment Programme). Fr. UnitedNations World Water Dev. Rep, Paris. 1–148 (2016).
Li, J. L. et al. Quality matters: pollution exacerbates water scarcity and sectoral output risks in China. Water Res. 224, 119059 (2022).
Ahsan, A., Ali, M. B. & Rashid, M. R. Sustainable leachate treatment by integrating electrolysis with palm-shell activated carbon contactor for environmental protection. Sci. Rep. 15, 560 (2025).
Jia, X. et al. Groundwater depletion and contamination: Spatial distribution of groundwater resources sustainability in China. Sci. Total Environ. 672, 551–562 (2019).
Pan, D., Chen, H., Zhang, N. & Kong, F. B. Do livestock environmental regulations reduce water pollution in China?. Ecol. Econ. 204, 107637 (2023).
Tao, T. & Xin, K. Public health: A sustainable plan for China’s drinking water. Nature. 511, 527–528 (2014).
Francois, D. & Youssef, Z. Cost-benefit analysis of nitrate abatement in the Souffel catchment (France): Sensitivity study of the damage and spatialization of the abatement effort. Environ. Impact. Assess. Rev. 95, 106791 (2022).
Wang, M. R., Bodirsky, B. L. & Rijneveld, R. A triple increase in global river basins with water scarcity due to future pollution. Nat. Commun. 15, 880 (2024).
Lu, J. Turnover of environmental protection officials and transboundary water pollution. Environ. Sci. Pollut. Res. 28, 10207–10223 (2023).
Sheng, J. C. & Webber, M. Incentive coordination for transboundary water pollution control: the case of the middle route of China’s South-North water transfer project. J. Hydrol. 598, 125705 (2021).
Zhang, Y. Y., Hu, N. Y., Yao, L. L., Zhu, Y. C. & Ma, Y. S. The role of social network embeddedness and collective efficacy in encouraging farmers’ participation in water environmental management. J. Environ. Manage. 340, 117959 (2023).
Yao, W. J. Effect of the River Chief System on river pollution control under responsibility contracting: Lessons and experiences from China. Water Policy. 25, 731–741 (2023).
Ding, R. & Sun, F. C. Impact of river chief system on green technology innovation: Empirical evidence from the Yangtze river Economic Belt. Sustainability. 15, 675 (2022).
Wang, J., Guo, X. N. & Jiang, Q. B. The river chief system and the total factor productivity in China: Evidence from the industrial enterprises database. Environ. Sci. Pollut. Res. 30, 50319–50331 (2023).
He, G., Wang, S. & Zhang, B. Watering down environmental regulation in China. Q. J. Econ. 135, 2135–2185 (2020).
Liu, Y. Z., Cheng, Y. Q., Li, T. S., Ni, J. L. & Norman, S. Information disclosure and public participation in environmental management: Evidence from the river chief system in China. China. Econ. Review. 85, 102168 (2024).
Shen, K. R. & Jin, G. The policy effects of the environmental governance of Chinese local governments: A study based on the progress of river chief system. Soc. Sci. China. 41, 87–105 (2020).
Yang, C. X. & Song, T. River chief governance in China: Trends and outlooks. Front. Environ. Sci. 12, 1396196 (2024).
Li, J., Shi, X., Wu, H. Q. & Liu, L. W. Trade-off between economic development and environmental governance in China: An analysis based on the effect of river chief system. China Econ. Rev. 60, 101403 (2020).
Mu, L., Zhang, C. & Liu, H. The impact of the river chief system on corporate ESG performance: Evidence from China. Water. 17, 265 (2025).
Vinca, A. et al. Transboundary cooperation a potential route to sustainable development in the Indus basin. Nat. Sustain. 4, 331–339 (2021).
Zheng, X., Tan, Y. & Li, D. S. Navigating environmental governance in China’s hog sector: unraveling the “race to the bottom” phenomenon and spatial dynamics. J. Knowl. Econ. https://doi.org/10.1007/s13132-024-01800-8 (2024).
Masina, F. M., Wasserman, R. J., Wu, N. C. & Mungenge, C. P. Macroinvertebrate diversity in relation to limnochemistry in an austral semi-arid transboundary aquifer region pan system. Sci. Total. Environ. 878, 163161 (2023).
Hu, C. Y., Zhou, F. J. & Li, W. (2025) How is holistic river water governance possible in China? The case of river chief system in Hunan province. J. Asian Pub. Policy https://doi.org/10.1080/17516234.2024.2448032 (2024).
UNSD (United Nations Statistics Division). Global Indicator Framework for the Sustainable Development Goals and Targets of the 2030 Agenda for Sustainable Development Report. (2017).
Chen, S. & Chen, D. Air pollution, government regulations and high-quality economic development. Econ. Res. J. 53, 20–34 (2018).
Kahn, M. E., Li, P. & Zhao, D. Water pollution progress at borders: The role of changes in China’s political promotion incentives. Am. Econ. J. Econ. Policy. 7, 223–242 (2015).
Koizumi, R., Kusama, Y. & Asai, Y. Effects of population age structure on parenteral antimicrobial use estimations. Sci. Rep. 13, 840 (2023).
Xu, X. S., Cheng, Y. Y. & Meng, X. C. River Chief System, emission abatement, and firms’ profits: Evidence from China’s polluting firms. Sustainability. 14, 3418 (2022).
Zhou, L., Li, L. Z. & Huang, J. K. The River Chief System and agricultural non-point source water pollution control in China. J. Integr. Agric. 20, 1382–1395 (2021).
Zhang, Z. X., Zou, X., Zhang, C. & Sharifi, S. Has China’s river chief system improved the quality of water environment? Take the Yellow River Basin as an example. P. J. Environ. Stud. 32, 4403–4416 (2023).
She, Y., Liu, Y., Jiang, L. & Yuan, H. Is China’s River Chief policy effective? Evidence from a quasi-natural experiment in the Yangtze River Economic Belt. China. J. Clean Prod. 220, 919–930 (2019).
Shen, K. & Jin, G. The policy effects of local governments’ environmental governance in China: A study based on the evolution of the river director system. Soc. Sci. China. 5(92), 115 (2018).
Jacobson, L. S., LaLonde, R. J. & Sullivan, D. G. Earnings Losses of Displaced Workers. Am. Econ. Rev. 83, 685–709 (1993).
Chetty, R., Looney, A. & Kroft, K. Salience and taxation: Theory and evidence. Am. Econ. Rev. 99, 1145–1177 (2009).
Moser, P. & Voena, A. Compulsory licensing: Evidence from the trading with the enemy act. Am. Econ. Rev. 102, 396–427 (2012).
Tang, C., Zhao, L. & Zhao, Z. Does free education help combat child labor? The effect of a free compulsory education reform in rural China. J. Popul. Econ. 33, 601–631 (2020).
Sun, D. Y., Wang, X. X., Yu, M. L., Ouyang, Z. L. & Liu, G. Dynamic evolution and decoupling analysis of agricultural nonpoint source pollution in Taihu Lake Basin during the urbanization process. Environ. Impact Assess. Rev. 100, 107048 (2023).
Li, T., Wang, Y. & Zhao, D. Environmental kuznets curve in China: New evidence from dynamic panel analysis. Energy Policy. 91, 138–147 (2016).
Heckman, J. J., Ichimura, H. & Todd, P. Matching as an econometric evaluation estimator. Rev. Evon. Stud. 65(261), 294 (1998).
Xiao, Y. X. & Nie, Y. Q. The technological advancement environmental regulations and their impact on energy efficiency and CO2 emissions. Sci. Rep. 15, 2781 (2025).
Agovino, M., Cerciello, M. & Musella, G. Economic cycle, labour market and pro-environmental behaviours. The case of separate waste collection in Italy. Environ. Impact Assess. Rev. 102, 107207 (2023).
Naqvi, R. A., Almohsen, B. & Sohail, A. Modeling the environmental Kuznets curve: A stochastic approach using economic and climate data. J. Environ. Manage. 373, 123108 (2025).
Gorus, M. S. Do tight environmental regulations cause economic contraction? Panel evidence from the European countries. Nat. Resour. Forum. https://doi.org/10.1111/1477-8947.12322 (2023).
Liu, X. H., Yang, J. J., Xu, C. Z., Li, X. C. & Zhu, Q. Y. Environmental regulation efficiency analysis by considering regional heterogeneity. Resour. Policy. 83, 103735 (2023).
Nguyen, K. T. Formal versus informal system to mitigate non-point source pollution: An experimental investigation. J. Agric. Econ. 71, 838–852 (2020).
Omoni, V. T., Bankole, P. O. & Omoche, O. Evaluation of the effects of abattoir effluent on the physicochemical and bacteriological quality of rive benue. Nigeria. Environ. Monit. Assess. 195, 146 (2023).
Greenstone, M. & Hanna, R. Environmental regulations, air and water pollution, and infant mortality in India. Am. Econ. Rev. 104, 3038–3072 (2014).
Bourceret, A., Amblard, L. & Mathias, J. D. How do farmers’ environmental preferences influence the efficiency of information instruments for water quality management? Evidence from a social-ecological agent-based model. Ecol. Model. 478, 110300 (2023).
Lipscomb, M. & Mobarak, A. M. Decentralization and pollution spillovers: Evidence from the re-drawing of county borders in Brazil. Rev. Econ. Stud. 84, 464–502 (2017).
Chakraborty, P., Chakrabarti, A. S. & Chatterjee, C. Cross-border environmental regulation and firm labor demand. J. Environ. Econ. Manag. 117, 102753 (2023).
Yin, L. & Wu, C. Promotion incentives and air pollution: From the political promotion tournament to the environment tournament. J. Enviro. Manag. 317, 115491 (2022).
Acknowledgements
This research was funded by the National Natural Science Foundation of China (42001221); Shaanxi Social Science Fund (2022D049); the major projects of the humanities and Social Sciences base of the Ministry of Education (22JJD790052); the third Xinjiang Scientific Expedition and Research Program (2022xjkk0305).
Funding
The National Natural Science Foundation of China,42001221.
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Mu, L., Zhang, C., Zeng, X. et al. The impact of the river chief system on transboundary water pollution. Sci Rep 15, 8192 (2025). https://doi.org/10.1038/s41598-025-92503-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-025-92503-w
