Impact of River Chief System on Green Technology Innovation: Empirical Evidence from the Yangtze River Economic Belt

Abstract

The River Chief System (RCS) is an innovative environmental governance system with Chinese characteristics that is significant for green and sustainable development, and green technology innovation (GTI) is a key step to achieve this goal. However, existing studies have not proved the effect of RCS on GTI. Therefore, this paper takes the implementation of RCS as a quasi-natural experiment and the progressive spatial difference-in-differences model is used to empirically investigate the effect of water environment governance policies on GTI, based on the panel data of 108 prefecture-level cities in the Yangtze River Economic Belt (YREB) from 2004 to 2019. The results of this research show that: (1) GTI in the YREB shows a rapid growth trend, and the lower reaches are generally higher than the middle and upper reaches; (2) RCS can improve the local GTI by 19.43% and has a significant positive incentive effect on adjacent regions’ GTI, while the GTI itself can generate a positive spillover effect for adjacent regions; (3) Heterogeneity analyses indicate that RCS has a stronger facilitation in the spontaneous parallel diffusion form cities and middle and lower reaches, while having an inhibition in riverine cities. In terms of spatial effects, RCS has a stronger positive spillover effect in the adjacent untreated area and upper reaches, while having a negative spillover effect in the spontaneous parallel diffusion form cities; (4) Government governance, official incentive and social supervision can enhance the effect of RCS on GTI. This study provides useful empirical evidence for environmental governance and green sustainable development of the YREB.

1. Introduction

With the rapid development of China’s economy, the balance between environmental protection and economic development has become a major contradiction for China, and green development has become a key to solve this dilemma. In March 2018, the “ecological civilization” was written into the Constitution for the first time, which was an important change in China’s economic development pattern, and it means that green development has entered a stage of normalization and institutionalization. How to achieve green development is an urgent issue for the Chinese government, and environmental regulation is one of the important ways to achieve green transformation development [1,2].

Faced with a complicated water pollution control problem, the Chinese government creatively proposed the “River Chief System” (RCS). RCS refers to the appointment of major party and government officials at all levels as “river chiefs”, who are responsible for organizing and leading the management and protection of rivers and lakes in their areas, so as to strengthen the prevention and control of water pollution and achieve effective management of the water environment. Different from the traditional mode of environmental governance, RCS creatively proposes that every river be guarded by a river chief. This chiefly responsible environmental governance model further enhances the political position of environmental governance, makes the environmental governance performance truly become the performance evaluation index of local party and government work, and makes for a great step forward in the process of ecological civilization construction. In this important stage of green transformation development of the Chinese economy, green innovation can break through the difficulty of environmental governance efforts and economic development efficiency to become an effective mechanism to realize green sustainable development [3]. As a form of innovation that is designed to protect the environment, green technology innovation (GTI) has become an important integration point of economic development and environmental protection, which can provide new momentum to break environmental constraints and promote high-quality development [4,5,6]. At present, the momentum of GTI in China is still insufficient. Therefore, whether the RCS can effectively stimulate GTI and provide long-term impetus for GTI becomes an urgent issue for us to study.

The Yangtze River Economic Belt (YREB) is densely populated with heavy polluting industries, and the problem of “chemical industry surrounding the Yangtze River” is prominent. The traditional extensive industrial development mode has exerted a bad impact on the ecological environment, severely restricting the green and sustainable development of the YREB. It is necessary to promote reform of the development mode and promote green transformation development of the YREB through GTI. Therefore, whether the RCS, a water environment management policy, can be a good medicine to solve environmental pollution and promote the GTI, needs to be explored in depth. Due to its complexity, diversity and uncertainty, the environmental governance of rivers and lakes presents the characteristics of “persistent disease” and has always been a difficult problem plaguing the world [7]. On the one hand, since the watershed is shared by multiple political units, the watershed environmental governance is more dependent on cross-regional cooperation [8]. However, the existence of information barriers in each region makes cross-regional cooperation become a major problem in watershed water environment governance. On the other hand, the cross-regional and multi-sector governance structure inevitably leads to the fragmentation of the water environment governance system, resulting in the problem of “Kowloon water control” [9]. In order to achieve effective management of the water environment, governments have made many attempts. Germany and France tend to adopt a centralized management model to control water environmental pollution in their watershed basins [10]. The United States adopts a decentralized governance model that delegates some environmental authority to the states [11]. China’s RCS has become a major institutional innovation in the water environment governance process through government regulation and non-governmental supervision. In 2007, in order to solve the cyanobacteria crisis in Taihu Lake, the Wuxi Municipal Government of Jiangsu Province tried RCS for the first time and achieved remarkable results. As a result, provinces and cities have followed the example of Wuxi to carry out water environment management in watershed basins. In December 2016, the General Office of the CPC Central Committee and the General Office of the State Council officially issued “the Opinions on the Comprehensive Implementation of the River Chief System”, which clearly proposed that the RCS should be fully established by the end of 2018; thus, China’s RCS began to enter the comprehensive implementation stage. The RCS is a water environment management system derived from the leading supervision system of river water quality improvement and the environmental protection accountability system. This policy delegates the responsibility of water environment management to local governments and incorporates environmental governance performance into a performance evaluation index. It breaks the boundaries of administrative regions and achieves the goal of water environmental governance by establishing an environmental governance mechanism of “joint prevention and control” [12]. As a hierarchical water environmental management policy, RCS has delegated the responsibility of river and lake water environmental management to all levels of government and departments [13], which effectively alleviates the collaborative problems in the process of watershed water environmental management through horizontal level and vertical level coordination mechanisms [14].

Most of the previous literature on the effects of RCS has focused on the pollution control effects of the RCS. For example, the implementation of RCS has significantly improved the efficiency of water environment governance, and can significantly reduce the level of environmental pollution, so as to significantly improve the quality of surface water [15,16,17]. A study of county data in China showed that the implementation of RCS can not only significantly reduce the negative impact of animal manure on surface water quality, but also effectively alleviate agricultural non-point source pollution [18]. Evidence from a micro perspective showed that the implementation of RCS can reduce an enterprises’ water pollution, improve the quality of water environment information disclosure of enterprises in water-sensitive industries, and enhance the enterprises’ green total factor energy efficiency [19,20,21]. With the increasing research on the effects of RCS, more and more scholars have begun to pay attention to the economic consequences brought by RCS. Although the RCS can obviously reduce the level of environmental pollution, it also has a short-term inhibitory effect on economic development [22,23]. In the new stage of China’s economic development, the problem to be solved by RCS is not only to reduce environmental pollution, but also to take into account its economic cost, so as to truly implement the concept of green development.

According to Schumpeter’s innovation theory, innovation is a new combination of production factors and production conditions, the purpose of which is to obtain potential profits and achieve economic development [24]. Different from traditional technological innovation that improves enterprise productivity and drives economic growth [25], the essence of GTI is to reduce environmental pollution and improve resource utilization efficiency, thus gaining competitive advantages and ultimately promoting green sustainable development [26]. Tao et al. [27] conducted a quasi-natural experiment on an environmental target responsibility system and found that environmental regulation can effectively improve the level of GTI. Wang et al. [28] took an environmental information announcement system as a quasi-natural experiment and found that an environmental information announcement could significantly stimulate green innovation. Lv et al. [29] studied Canada’s environmental policies and found that strict environmental policies have a positive impact in promoting enterprise innovation, while relaxed environmental policies reduce enterprise innovation. Feng et al. [30] found that empowering local governments to conduct environmental governance can help enhance the ability of digital finance to support enterprises’ GTI activities. In a study of listed companies in China, Cui et al. [31] found that cleaner production audit programs significantly promoted the development of green patents. More interestingly, CPA schemes have a stronger positive effect on radical environmental innovation than incremental environmental innovation. Cai et al. [32] found that direct environmental regulation has a strong and significant incentive effect on green technology innovation in heavily polluting industries.

As mentioned in the literature, the implementation of RCS is crucial to green sustainable development, and GTI is a key catalyst to promote sustainable development. Therefore, in order to deeply understand the effects of RCS, it is necessary to explore whether RCS can stimulate GTI. Unfortunately, few studies have been concerned with this issue. Therefore, this paper adopts the SDID method to evaluate the effects of RCS on GTI. Compared with the existing literature, this paper makes the following contributions. Firstly, and differently from the existing literature that focuses on the effects of RCS on pollution control, this paper focuses on the impact of RCS on green sustainable development, deconstructs the green incentive effect of RCS from the perspective of cross-regional communication, and estimates the spatial spillover effect of RCS on GTI, all of which broadens the dimensions of this study on the effects of RCS. Secondly, we discuss the theoretical mechanism of RCS affecting GTI from the three aspects of government governance, official incentives, and social supervision, which provides a new direction for green development of YREB and complements the literature on the effects of RCS. Thirdly, we fully explore the heterogeneous effects of RCS on GTI from the perspective of policy implementation differences and regional differences. Different characteristics and forms of action of policies will produce different implementation effects, and the effects of policies in different regions will also be different. The heterogeneity analysis is conducive to improving the applicability of RCS in different situations. This study has important implications for water environment governance and green sustainable development in YREB.

The remainder of this paper is organized as follows. Section 2 is the mechanism analysis. Section 3 describes the methodology and data used in the paper. The empirical results and analysis are reported in Section 4. The final section provides conclusions and policy implications.

2. Mechanism Analysis

2.1. The Direct Effect of RCS on GTI

(1)

Government governance effect. YREB is the main force of China’s economic development, among which the upper reaches still face heavy tasks of economic development. The implementation of RCS makes the contradiction between economy and environment in YREB more and more prominent. As an important driving force of economic development, GTI has become the focus of local governments in the governing process. RCS is an environmental decentralization method that can empower local governments with greater environmental management powers and give full play to their subjective initiatives for environmental management, thus effectively promoting GTI [33]. The clarity of local governments’ environmental governance responsibilities and the strengthening of environmental supervision can form a “race to the top” environmental control mechanism among regions and urge enterprises to carry out GTI [34]. RCS puts forward clear requirements for water pollution prevention and control, requiring governments to strictly implement the “Action Plan for Water Pollution Prevention and Control” and strictly control industrial pollution discharge. On the one hand, raising the emission tax standard, and using a policy punishment mechanism to force polluters to carry out green technology innovation and transformation. On the other hand, increasing the financial support for green innovation, so as to encourage enterprises to continue to carry out GTI.

(2)

Official incentive effect. The proposal of major decisions in a region affects the transformation of its development mode. RCS is a concrete incentive for local governments to take responsibility for the water environment quality inside their jurisdictions, so its implementation effect largely depends on whether the local party and government leaders pay sufficient attention to environmental governance. The political will and commitment of local officials can provide a powerful driving force for local environmental governance [35]. According to RCS, four levels of river chiefs are set up, including provincial, city, county, and township. Major party and government leaders at all levels are appointed as river chiefs, and comprehensive evaluation is conducted on leading officials based on an audit of their natural resources and assets when they leave office. The comprehensive implementation of RCS has put forward new requirements for the performance assessment of local officials, and further strengthened the “green GDP” assessment of local government officials through the government accountability mechanism and social supervision mechanism, so as to encourage local officials to strictly implement the provisions of RCS. The goal-oriented performance evaluation system of environmental governance provides powerful political incentives for local officials to reduce pollution [36]. The strict “one vote veto” evaluation and accountability mechanism of RCS encourages local officials to give up window-dressing pollution treatment, strictly carry out environmental management, and effectively improve the intensity of regional environmental regulation. According to Porter’s hypothesis, appropriate environmental regulations can stimulate the innovation vitality of enterprises and generate “innovation compensation effect”, whose economic benefits can offset environmental protection costs and improve production efficiency [37], so as to encourage enterprises to develop green technologies to reduce pollution control costs.

(3)

Social supervision effect. Different from other environmental governance policies, RCS expands the public participation mode from the traditional government governance mode and sets the civil river chiefs to participate in river environmental governance. Encouraging the masses to participate in and supervise the work of RCS helps to realize the water environment governance in which the whole populace participates. Public participation is essential to promote effective environmental governance [38]. The public supervision mechanism of RCS makes the public really participate in pollution control, so as to further enhance the enthusiasm of the public in pollution control. On the one hand, the public supervises and reports illegal emission behaviors through petition platforms, the internet, telephone, and other channels, so as to urge the government to carry out effective environmental control, thus forcing enterprises to strengthen pollution control and carry out GTI. On the other hand, with the further deepening of the public’s awareness of environmental protection and the improvement of the enthusiasm for environmental governance, as well as the public’s expectation of beautiful rivers and lakes and the demand for a better life, when shopping people would like to choose clean products and green products. Based on supply and demand theory, enterprises will be incentivized to carry out GTI and produce pollution-free green products to meet market demand. Green products that generate more economic benefits are more easily accepted by the public, and as a result, more enterprises are encouraged to carry out GTI. The specific mechanism analysis is shown in Figure 1. Based on the above analysis, we propose H1 as follows:

Hypothesis 1.

The RCS can significantly promote urban GTI.

2.2. The Spatial Spillover Effect of RCS on GTI

According to the theory of new economic geography, the proximity and the development differences between different regions make regional synergistic development become affected by spatial spillover effects. Due to the interconnectivity between regions, the policies implemented in one region can have an impact on the development strategies of other regions. Therefore, because of the existence of spatial interaction, when a certain region implements RCS, other regions will carry out strategic imitation and promote the implementation of RCS in neighboring areas through demonstration effect, thus having a positive effect on GTI in neighboring areas.

On the other hand, GTI itself also has spatial spillover effects. According to the theory of factor flow, the inter-regional flow of factors will form polarization, diffusion, and injection effects. Due to the unbalanced distribution of resources and the imbalance of regional development, this results in the flow of factors, and forms the exchange of capital and labor among regions. As an open system, the flow of capital and labor force will make innovation technology spread among regions. According to the knowledge spillover theory, the difference in innovation level will affect the factor allocation efficiency, resulting in the dynamic flow of innovation resources among different regions, forming a positive spatial spillover effect of GTI. To sum up, the RCS will have a positive spillover effect on GTI in the neighboring areas. Based on the above analysis, we propose H2 as follows:

Hypothesis 2.

The RCS can produce positive spatial spillover effects on GTI in neighboring areas.

3. Methodology and Data

3.1. Econometric Model

3.1.1. The Difference-In-Differences (DID) Model

The DID method is often used to evaluate policy effect; its advantage is that it can largely avoid the problems of endogeneity. In order to estimate the impact of RCS on GTI in YREB, this paper follows Beck et al. [39] and constructs the progressive DID regression model as follows:

$$GT{I}_{it}={\beta }_0+{\beta }_{1}RC{S}_{it}+\lambda X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(1)

where subscripts i and t denote city and year and GTI_it represents the green technology innovation of the city i in year t. RCS_it is a dummy variable that equals 1 in the years after city i has initiated RCS and 0 otherwise. β₁ indicates the influence of RCS on GTI. X_it represents a series of control variables affecting green technology innovation, including financial development (FD), industrial structure (IS), level of opening-up (Open), regional education level (Edu), infrastructure (Rod), and government science and technology support (GS). μ_it and ν_it are vectors of the city and year dummy variables that indicate city and year fixed effects, respectively, and ε_it is the random error term.

3.1.2. The Spatial Difference-In-Differences (SDID) Model

Although the DID model with exogenous policy as a quasi-natural experiment can alleviate the endogeneity problems caused by bidirectional causality and omitted variables, the traditional DID model fails to take into account the spatial correlation of policy impact, which violates the stable unit processing value hypothesis (SUTVA) and makes it impossible to obtain unbiased estimation [40]. The SDID model can solve this problem so in order to accurately estimate the impact of RCS on GTI in YREB, we follow Li and Luo [41] in constructing the SDID model based on the traditional DID model. Since GTI was affected by RCS of local and neighboring areas, as well as GTI in neighboring areas, we follow Elhorst [42] in using the spatial Dubin model (SDM) to test the spatial spillover effect of RCS. The form of SDM can be described as follows:

$$GT{I}_{it}={\beta }_0+\rho W\times GT{I}_{it}+{\beta }_{1}RC{S}_{it}+\lambda X_{it}+\theta W\times RC{S}_{it}+\gamma W\times X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(2)

where W indicates spatial weight matrix and ρ, θ and γ represent the spatial effect estimation coefficient. Other parameters are the same as in Model (1). In order to effectively avoid the endogenous problem, we use the maximum likelihood estimation (MLE) to estimate the SDM.

3.1.3. Spatial Weight Matrix

The RCS is a watershed environmental governance system; due to the transboundary characteristics of the watershed, the implementation of the policy is more sensitive to its adjacent areas. Therefore, we construct the geographical adjacency spatial weight matrix (W) to represent the spatial correlation. W_ij is the element of the spatial weight matrix and the format is as follows:

$$W_{ij}=\{\begin{array}{cc}1,\hfill & \mathit{City\; i\; and\; j\; are\; adjacent}\hfill \\ 0,& \mathit{City\; i\; and\; j\; are\; not\; adjacent}\hfill \end{array}$$

(3)

3.2. Variables Definitions

3.2.1. Explained Variable

The explained variable is green technology innovation (GTI). GTI is measured by the sum of the number of green invention patents and green utility model patents granted, and the natural logarithm of the number of green patents is obtained. The number of green patents granted was obtained from the Green Patent Research Database in the China Research Data Service Platform.

3.2.2. Explanatory Variable

The explanatory variable is the River Chief System (RCS). We use dummy variables to indicate whether prefecture-level cities implement RCS. The value is 1 if a city has implemented the RCS for the year, and 0 if otherwise. The data on the implementation of the RCS of each prefecture-level city was obtained from the China National Knowledge Internet, Peking University Law Database, and People’s Government Network of each prefecture-level city. We searched the keywords “River Chief System“ on the website and collected information on the implementation of RCS in each prefecture-level city.

3.2.3. Control Variable

(1) Financial development (FD), which takes the ratio of the sum of loan balances and deposit balances of financial institutions in each region to GDP as a proxy variable. (2) Industrial structure (IS), measured by the proportion of the output value of the tertiary industry to the output value of the secondary industry. (3) Level of opening-up (Open), measured by the proportion of the actual utilization of foreign direct investment in GDP of each city, and is converted by the average price of RMB exchange rate over the years. (4) Regional education level (Edu), measured by the natural logarithm of the number of regional higher education institutions. (5) Infrastructure (Rod), measured by the area of urban roads occupied per capita. (6) Government science and technology support (GS), measured by the proportion of science and technology expenditure in local fiscal budget expenditures.

3.3. Data Sources

This paper selects the panel data of 108 prefecture-level and above cities in the YREB from 2004 to 2019 as the research sample. Since Tongren and Bijie cities were newly established as prefecture-level cities in 2011, they were excluded from this study. In addition to explanatory variables and explained variables, other data required for this article were obtained from the China Urban Statistical Yearbook, the China Regional Economic Statistical Yearbook, and the statistical yearbooks and national economic development statistical bulletins of 108 prefecture-level cities in the YREB. The few missing data points were completed by linear interpolation. Descriptive statistics for the main variables are shown in

Table 1. Descriptive statistics of variables.

Variables	Obs	Mean	Median	St. Dev	Min	Max
GTI	1728	4.0447	3.9796	1.8948	0.0000	9.1913
RCS	1728	0.2870	0.0000	0.4525	0.0000	1.0000
FD	1728	2.1556	1.9045	0.8998	0.7637	6.4347
IS	1728	0.8687	0.8060	0.3771	0.2723	4.9308
Open	1728	0.0225	0.0166	0.0202	0.0000	0.1207
Edu	1728	1.7140	1.4979	0.8928	0.0000	4.4427
Rod	1728	3.5117	2.5464	3.1489	0.0440	22.9775
GS	1728	0.0161	0.0105	0.0166	0.0000	0.1627

.

4. Empirical Results and Analysis

4.1. Non-Spatial DID Model Analysis

Table 2. Non-spatial DID model regression results.

Model	(1)		(2)
Variable	Coefficient	Standard-Error	Coefficient	Standard-Error
RCS	2.0489 ***	(0.0537)	0.0779 *	(0.0444)
FD			0.1747 ***	(0.0404)
IS			−0.0924 *	(0.0558)
Open			0.4649	(0.9092)
Edu			0.0791	(0.0554)
Rod			0.0316 ***	(0.0095)
GS			11.5569 ***	(1.1108)
Cons	3.4566 ***	(0.0277)	1.6273 ***	(0.1270)
City Effect	Yes		Yes
Year Effect	Yes		Yes
N	1728		1728
R2	0.4731		0.9123
Log-like	−2318.4676		−769.1677
AIC	4640.9351		1584.3355
BIC	4651.8445		1709.7940

Note: * and *** indicate significance at the 10% and 1% levels, respectively.

shows the influence of RCS on GTI without spatial effect. As shown in Table 2, the coefficient of RCS is 2.0489 when no control variables are added, and it is significant at the 1% level. After adding the control variables, although the coefficient and significance are reduced, it is still significantly positive, which indicates that the implementation of RCS can effectively stimulate GTI in the YREB.

4.2. Parallel Trend Test

In order to estimate the real policy effect, the treatment group and the control group need to satisfy the parallel trend assumption before the policy is implemented. In this paper, we follow Jacobson et al. [43] by adopting the event study method to test the parallel trend, and to estimate the dynamic effects of policies over time. The test model is as follows:

$$GT{I}_{it}={\beta }_0+{\beta }_t{\displaystyle {\sum }_{t=-m}^{n}Trea{t}_{i}Yea{r}_t}+\lambda X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}$$

(4)

where Year_t denotes the year dummy variables before and after the policy implementation, the starting year of RCS is used as the base group, and the dummy variables for the 5 years before and 5 years after the policy implementation are set separately to generate the interaction term between the year dummy variables and the treatment group dummy variables to test the dynamic effect of RCS. If the coefficient β_t is not significant before the policy implementation, it indicates that the parallel trend hypothesis is fitted. The results in

Table 3. Parallel trend test.

Variable	Coefficient/Standard-Error	Variable	Coefficient/Standard-Error
t_5	−0.0017	t1	0.0277
	(0.0475)		(0.0546)
t_4	0.0153	t2	0.0310
	(0.0489)		(0.0541)
t_3	0.0356	t3	0.1769 ***
	(0.0543)		(0.0638)
t_2	0.0236	t4	0.1938 ***
	(0.0510)		(0.0714)
t_1	0.0130	t5	0.1106 *
	(0.0574)		(0.0650)
t0	0.0589	Cons	1.6205 ***
	(0.0583)		(0.2431)
Control Variables	Yes
City Effect	Yes
Year Effect	Yes
N	1728
R2	0.913

Note: Values in parentheses are standard error; * and *** indicate significance at the 10% and 1% levels, respectively.

show that none of the coefficients β_t are significant before the implementation of RCS, indicating that the treatment and control groups satisfy the parallel trend hypothesis. Figure 2 plots the dynamic effect of RCS at 90% confidence interval. After the policy is implemented, the coefficient β_t is not significant in the first two years, and from the third year, the coefficient is significantly positive and shows an inverted U-shaped trend of rising and then falling over time, indicating that the effect of RCS on GTI has a lag.

4.3. Robustness Tests

In order to ensure the reliability of the test results, this paper conducts the following robustness tests: (1) Placebo test. A number of contingent factors accompanying the implementation of RCS may also influence GTI, and to exclude the possibility of these effects, this paper follows Abadie and Gardeazabal’s [44] study by using counterfactual events for placebo testing. Advancing the time of RCS by 1–3 years, respectively, to generate false policy variables (Falsepolicy) is undertaken before putting Falsepolicy into model (1) for estimation. (2) Replacing the index of explained variable. The number of green invention patent and green utility model patent applications were used as a proxy variable for GTI to estimation.

Table 4. Robustness test.

Variable	(1)	(2)	(3)	(4)
RCS				0.0895 **
				(0.0414)
Falsepolicy	0.0455	0.0421	0.0540
	(0.0440)	(0.0436)	(0.0433)
FD	0.1756 ***	0.1751 ***	0.1764 ***	0.1828 ***
	(0.0405)	(0.0405)	(0.0404)	(0.0377)
IS	−0.0853	−0.0837	−0.0835	−0.2767 ***
	(0.0558)	(0.0556)	(0.0554)	(0.0521)
Open	0.3443	0.2898	0.2164	2.5139 ***
	(0.9058)	(0.9041)	(0.9056)	(0.8482)
Edu	0.0752	0.0748	0.0744	0.0145
	(0.0554)	(0.0554)	(0.0554)	(0.0517)
Rod	0.0328 ***	0.0332 ***	0.0338 ***	0.0179 **
	(0.0095)	(0.0095)	(0.0095)	(0.0089)
GS	11.5941 ***	11.6205 ***	11.6014 ***	11.2239 ***
	(1.1128)	(1.1113)	(1.1110)	(1.0362)
Cons	1.6260 ***	1.6268 ***	1.6249 ***	2.2090 ***
	(0.1272)	(0.1273)	(0.1270)	(0.1184)
City Effect	Yes	Yes	Yes	Yes
Year Effect	Yes	Yes	Yes	Yes
N	1728	1728	1728	1728
R2	0.9122	0.9122	0.9122	0.9302
AIC	1586.5059	1586.6553	1585.9860	1344.2499
BIC	1711.9644	1712.1139	1711.4445	1469.7085

Note: Values in parentheses are standard error; ** and *** indicate significance at the 5% and 1% levels, respectively.

reports the results of the robustness test and columns (1)–(3) show the estimated results of advancing the policy time by 1–3 years. After advancing the policy time, the coefficient of Falsepolicy is no longer significant, indicating that GTI is promoted by RCS and the results of this paper are robust. Column (4) shows the estimated result of replacing the explained variable whereby after replacing the explained variables, the RCS coefficient remains significant, which further indicates that the conclusion of this paper is robust.

4.4. The Temporal and Spatial Evolution of GTI

To demonstrate the temporal and spatial evolution characteristics of GTI in the YREB objectively and clearly, this paper uses ArcGIS software to map out the spatial distribution of GTI in the YREB. Figure 3 shows the spatial distribution of GTI in the YREB in 2004, 2009, 2014, and 2019. From 2004 to 2019, the level of GTI in the YREB has been greatly improved, with the number of green patents granted increasing from 2421 to 99,283 in 2004–2019, with an average annual growth rate of 267%. Among them, the downstream regions have a faster growth rate, with an average annual growth rate of 586%, while the middle and upstream regions have a slower development rate of green innovation. Moreover, GTI in different regions presents obvious spatial differences. In 2019, the level of GTI in downstream regions was generally higher than that in the middle and upstream regions, showing a high-level development trend on the whole, and showing a circle structure distribution spreading from central cities to surrounding cities. This means that GTI in downstream regions generates spillover dividends to surrounding regions. In the middle and upstream regions, except for Changsha, Wuhan, Chongqing, Chengdu, Guiyang, and other provincial capitals, the GTI of other cities is still at a low level, which is mainly due to the resource advantages of provincial capitals. The provincial capital city is generally the center of political, economic, and social development of a province (autonomous region), and all kinds of resources will gather in the provincial capital city, so that it has better development advantages. From the perspective of the spatial evolution model of GTI, the GTI in the YREB shows a certain spatial correlation, and the downstream regions show convergence characteristics.

4.5. Spatial Correlation Analysis

To comprehensively examine the spatial spillover effects of GTI in the YREB, exploratory spatial data analysis (ESDA) was used to test its spatial correlation.

Table 5. Global Moran’s I of GTI from 2004 to 2019.

Year	2004	2005	2006	2007	2008	2009	2010	2011
Moran’s I	0.1750 ***	0.2390 ***	0.2680 ***	0.3070 ***	0.3670 ***	0.3710 ***	0.3660 ***	0.3890 ***
Z Value	2.7130	3.6600	4.0740	4.6580	5.5260	5.5860	5.5160	5.8500
Year	2012	2013	2014	2015	2016	2017	2018	2019
Moran’s I	0.4110 ***	0.4160 ***	0.4580 ***	0.4230 ***	0.4740 ***	0.3930 ***	0.4380 ***	0.4010 ***
Z Value	6.1730	6.2460	6.8640	6.3450	7.1070	5.9060	6.5550	6.0310

Note: *** denotes significance levels of 1%.

reports the results of global Moran’s I of GTI in 108 cities in the YREB from 2004 to 2019, tested by a spatial adjacency matrix. As can be seen from Table 5, Moran’s I were positive in all years and passed the 1% significance test, indicating that there was a significant positive spatial correlation of GTI in the YREB. Therefore, it is necessary to discuss the effect of RCS on GTI from a spatial perspective.

4.6. Spatial Spillover Effects of RCS on GTI

The estimated results of Model (1) have proved that without introducing the spatial effect, the implementation of RCS can significantly increase the level of GTI in the YREB. In order to further verify whether there is a spatial spillover effect in the impact of RCS on GTI, this paper conducts a spatial effect estimation based on Model (2).

Table 6. Spatial effects of RCS on GTI.

Variable	Direct Effect	Indirect Effect	Total Effect
RCS	0.1943 ***	0.3846 ***	0.5788 ***
	(0.0504)	(0.0943)	(0.0965)
FD	0.2264 ***	0.9179 ***	1.1442 ***
	(0.0439)	(0.1029)	(0.1073)
IS	−0.2625 ***	−1.0157 ***	−1.2782 ***
	(0.0583)	(0.1379)	(0.1450)
Open	3.9775 ***	−0.2487	3.7289
	(1.0329)	(2.6935)	(2.8605)
Edu	0.3430 ***	1.5013 ***	1.8443 ***
	(0.0598)	(0.1928)	(0.2186)
Rod	0.1098 ***	0.2047 ***	0.3145 ***
	(0.0103)	(0.0314)	(0.0358)
GS	7.7905 ***	28.6014 ***	36.3919 ***
	(1.3655)	(3.1165)	(3.2685)
ρ	0.4490 ***
	(0.0208)
σ2	0.1766 ***
	(0.0061)
City Effect	Yes
Year Effect	Yes
N	1728
R2	0.8668
Log-like	−1002.0612
AIC	2064.1223
BIC	2227.7639

Note: Values in parentheses are standard error; *** indicates significance at the 1% level.

shows the results of the spatial effect of RCS on GTI. As shown in Table 6, the direct effect of RCS on GTI is 0.1943, indicating that RCS can improve the local GTI by 19.43%, which is significantly higher than that of the non-spatial model. It indicates that the introduction of spatial effect enhances the impact of RCS on GTI, and further indicates the necessity of introducing spatial correlation.

The indirect effect of RCS on GTI is 0.3846 and is significant at the level of 1%, which indicates that RCS can not only promote the local GTI, but also significantly improve GTI in neighboring areas. At the same time, the coefficient ρ is 0.4490 and is significant at the level of 1%, indicating that regional GTI itself can generate positive spillover effects to adjacent regions. According to the “iceberg transportation cost” model, the greater the spatial damping is, the more difficult it is for the green innovation technology to be applied by other regions. Compared with more distant regions, the spatial damping of adjacent regions is smaller, so green technology innovation is more likely to produce spatial spillovers.

4.7. Heterogeneity Analysis

The characteristics and the specific situation of the policy implementation are crucial to the policy effect [45], so it is necessary to conduct an in-depth analysis of the difference in the effect of RCS. This paper will explore the effect difference of RCS on GTI from the perspective of policy difference and geographical location.

4.7.1. Heterogeneity Analysis of Spatial Spillover Direction of RCS

In Model (2), the coefficient θ represents the indirect effect of RCS on GTI in neighboring areas, which is usually ignored in the traditional DID model. However, another important problem is that the indirect effect is the average effect of RCS on both treated and untreated areas but fails to separate its effect on different types of areas. Based on this, we follow the study of Chagas [40] by decomposing the spatial weight matrix as follows:

$$W=W_{T,T}+W_{T,NT}+W_{NT,T}+W_{NT,NT}$$

(5)

$$W_{T,T}=D_t\times W\times D_t$$

(6)

$$W_{T,NT}=D_t\times W\times D_t^c$$

(7)

$$W_{NT,T}=D_t^c\times W\times D_t$$

(8)

$$W_{NT,NT}=D_t^c\times W\times D_t^c$$

(9)

$$D_t=diag(T_t)$$

(10)

$$D_t^c=diag(I_n-T_t)$$

(11)

$$T_t=(T_1,T_2,\cdots ,T_t)\prime$$

(12)

where T_t is a dummy variable indicating treated regions. I_n is a vector of 1′s. W_ij represents the neighborhood effects of the j-region on i-region, i, j = T (treated) or NT (untreated), model (13) is obtained:

$$\begin{array}{c}GT{I}_{it}={\beta }_0+\rho W\times GT{I}_{it}+[\alpha +({\theta }_1{W}_{T,T}+{\theta }_2{W}_{NT,T}+{\theta }_3{W}_{T,NT}+{\theta }_4{W}_{NT,NT})]\times RC{S}_{it}\\ {\beta }_{1}RC{S}_{it}+{\lambda }_i{X}_{it}+{\gamma }_{i}W\times X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}\end{array}$$

(13)

Due to W_T,NTRCS_t and W_NT,NTRCS_t are 0-vectors (For the specific process, please refer to the study of Chagas (2016) [40]). The following model (14) is constructed to estimate the decomposition effect of RCS in different policy regions.

$$\begin{array}{c}GT{I}_{it}={\beta }_0+\rho W\times GT{I}_{it}+{\beta }_{1}RC{S}_{it}+{\lambda }_i{X}_{it}+[\alpha +({\theta }_1{W}_{T,T}+{\theta }_2{W}_{NT,T})]\times RC{S}_{it}\\ +{\gamma }_{i}W\times X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}\end{array}$$

(14)

where θ₁ represents the indirect effect of RCS on GTI in the adjacent treated area, and θ₂ represents the indirect effect of RCS on GTI in the adjacent untreated area. W_T,T and W_NT,T are spatial weight matrixes obtained by decomposition of W, representing the spatial relationship of the influence of the treated area on the treated area and the untreated area, respectively. Other parameters are the same as in Model (2).

Table 7. Heterogeneity results of spatial spillover direction.

Variable	Coefficient	Standard-Error
RCS	0.2566 ***	(0.0569)
WT,T × RCS	0.2479 ***	(0.0681)
WNT,T × RCS	0.3792 ***	(0.0633)
FD	0.1428 ***	(0.0472)
IS	−0.2034 ***	(0.0629)
Open	3.9107 ***	(1.0803)
Edu	0.2145 ***	(0.0587)
Rod	0.0865 ***	(0.0100)
GS	5.8842 ***	(1.3500)
W × FD	0.4291 ***	(0.0703)
W × IS	−0.5311 ***	(0.0931)
W × Open	−2.8327	(1.7849)
W × Edu	0.7003 ***	(0.1140)
W × Rod	0.0763 ***	(0.0195)
W × GS	15.3688 ***	(2.1046)
ρ	0.4240 ***	(0.0212)
σ2	0.1741 ***	(0.0060)
City Effect	Yes
Year Effect	Yes
N	1728
R2	0.8738
Log-like	−983.9483
AIC	2001.8965
BIC	2094.6268

Note: *** indicates significance at the 1% level.

shows the heterogeneity results of spatial spillover direction of RCS on GTI. After decomposing the spatial spillover effect, it is found that the effect of RCS on local GTI has been significantly improved (0.2566/0.1943), indicating that the general SDID model underestimates the policy effect of RCS. Therefore, when spatial effects are introduced, the policy effect on treated and untreated areas should be estimated separately.

The decomposition effect estimation results show that the impact of RCS on GTI of the neighboring treated areas is 0.2479, while the impact on the neighboring untreated areas is 0.3792, and both are significant at the level of 1%. It means that the spillover effect of RCS on GTI in the neighboring regions that have not implemented the policy is stronger than those that have implemented the policy. It indicates that in the process of environmental governance policy dissemination, regions are more willing to try new policies from other regions and implement new policies in their own regions through the imitation effect and the “learning by doing” effect, so as to improve the level of local environmental governance and green development. It so happens that the spillover effect of the policy decreases when all regions implement RCS. It also indicates that in the process of environmental governance, regions are paying more attention to new policies and less effort to further enhance the effects of old policies. Therefore, when all regions have realized RCS, efforts should be made to improve the effectiveness and adaptability of RCS, so as to realize the transformation of RCS from weak to strong.

4.7.2. Heterogeneity Analysis of Diffusion Forms of RCS

The RCS was first initiated by Wuxi City in Jiangsu Province in 2007. Subsequently, various regions followed the experience of Wuxi City to implement RCS to improve the water environment in watershed basins while forming the spontaneous parallel diffusion pattern of RCS. In 2016, the General Office of the CPC Central Committee and the General Office of the State Council officially issued the Opinions on the Comprehensive Implementation of RCS, requiring the comprehensive implementation of RCS governance mode nationwide, forming a vertical and downward diffusion mode of RCS. In the specific implementation and promotion of the policy, there are differences in the implementation motives of local governments under different diffusion modes, which will lead to different implementation effects of RCS [22]. Based on this, we will construct the following econometric model to estimate the governance effect of RCS under different diffusion modes [46,47].

$$\begin{array}{c}GT{I}_{it}={\beta }_0+\rho W\times GT{I}_{it}+{\beta }_{1}RC{S}_{it}+{\beta }_{2}RC{S}_{it}\times Typ{e}_i+{\lambda }_i{X}_{it}+{\theta }_{1}W\times RC{S}_{it}\\ +{\theta }_{2}W\times RC{S}_{it}\times Typ{e}_i+{\gamma }_{i}W\times X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}\end{array}$$

(15)

where Type_i is a dummy variable indicating the policy diffusion form, which is 1 when the policy diffusion form is spontaneous parallel diffusion, and 0 otherwise. Other variables are the same as Model (2).

Table 8. Heterogeneity results of diffusion forms.

Variable	Direct Effect	Indirect Effect
RCS	0.0547	0.5603 ***
	(0.0688)	(0.1061)
RCS × Type	0.1775 ***	−0.3108 **
	(0.0650)	(0.1248)
FD	0.2254 ***	0.9406 ***
	(0.0444)	(0.1034)
IS	−0.2561 ***	−1.0292 ***
	(0.0598)	(0.1448)
Open	4.4548 ***	−1.7545
	(1.0698)	(2.7443)
Edu	0.3436 ***	1.4926 ***
	(0.0611)	(0.1893)
Rod	0.1079 ***	0.2046 ***
	(0.0108)	(0.0309)
GS	7.3555 ***	30.2962 ***
	(1.2229)	(3.1191)
ρ	0.4453 ***
	(0.0209)
σ2	0.1756 ***
	(0.0061)
City Effect	Yes
Year Effect	Yes
N	1728
R2	0.8681
Log-like	−996.0718
AIC	2060.1436
BIC	2245.6040

Note: Values in parentheses are standard error; ** and *** indicate significance at the 5% and 1% levels, respectively.

shows the heterogeneity results of different diffusion forms of RCS. As shown in Table 8, the direct effect of the interaction terms of RCS and Type is 0.1775 and significant at the level of 1%, indicating that compared with the vertical diffusion form, the spontaneous parallel diffusion form makes the policy incentive effect of RCS increase by 17.75%. The parallel diffusion pattern of RCS is the spontaneous environmental governance behavior of local governments. There are generally two motivations for this behavior. First, there is serious water environmental pollution within the jurisdiction, which makes the urgency of environmental governance strong. Second, local governments have a strong responsibility for environmental governance and pay great attention to environmental governance. In the above two cases, in order to improve regional environmental quality, local governments will actively implement RCS and force enterprises to carry out GTI by raising the threshold of environmental access and pollution discharge standards, so as to generate innovation incentive effect.

In addition, the spatial spillover effect of the interaction terms is significantly negative, indicating that the spontaneous parallel diffusion form of RCS impedes the development of GTI in neighboring areas, and generates an “innovation crowding out effect”. The regions that voluntarily implement RCS will increase the intensity of environmental regulation while making the heavy polluting industries with low levels of GTI move to the surrounding regions that have not yet implemented RCS and have a low intensity of environmental regulation, which will have a negative impact on the GTI of the surrounding regions.

4.7.3. Heterogeneity Analysis of Geographical Location

As a water environment management policy, does the policy effect of RCS have obvious differences between the riverine and non-riverine cities? To answer this question, we built model (16) to estimate the difference in the effect of the RCS in riverine and non-riverine cities.

$$\begin{array}{c}GT{I}_{it}={\beta }_0+\rho W\times GT{I}_{it}+{\beta }_{1}RC{S}_{it}+{\beta }_{2}RC{S}_{it}\times Y{J}_i+{\lambda }_i{X}_{it}+{\theta }_{1}W\times RC{S}_{it}\\ +{\theta }_{2}W\times RC{S}_{it}\times Y{J}_i+{\gamma }_{i}W\times X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}\end{array}$$

(16)

where YJ_it is a dummy variable representing the nature of a city, YJ_it is 1 when the city is along the river, and 0 if otherwise. Other variables are the same as Model (2).

Based on the geographical location map of the YREB, among the 108 cities in the YREB, the main stream of the Yangtze River flows directly through 30 cities, including Lijiang, Kunming, Zhaotong, Panzhihua, Yibin, Luzhou, Chongqing, Yichang, Jingzhou, Yueyang, Xianning, Wuhan, Ezhou, Huanggang, Huangshi, Jiujiang, Anqing, Chizhou, Tongling, Wuhu, Ma ‘anshan, Nanjing, Zhenjiang, Yangzhou, Changzhou, Taizhou, Wuxi, Tongzhou, Suzhou, and Shanghai, defining these cities as riverine cities.

Table 9. Heterogeneity results of geographical location.

Variable	Direct Effect	Indirect Effect
RCS	0.2885 ***	0.3691 ***
	(0.0563)	(0.1058)
RCS × YJ	−0.2695 ***	0.0302
	(0.0654)	(0.1611)
FD	0.2224 ***	0.8620 ***
	(0.0443)	(0.1019)
IS	−0.2588 ***	−1.0569 ***
	(0.0598)	(0.1463)
Open	3.4099 ***	−1.1487
	(1.0585)	(2.7439)
Edu	0.3277 ***	1.4544 ***
	(0.0615)	(0.1932)
Rod	0.1117 ***	0.2145 ***
	(0.0108)	(0.0315)
GS	7.8558 ***	29.7100 ***
	(1.2110)	(3.1303)
ρ	0.4513 ***
	(0.0208)
σ2	0.1748 ***
	(0.0060)
City Effect	Yes
Year Effect	Yes
N	1728
R2	0.8679
Log-like	−993.7494
AIC	2055.4988
BIC	2240.9593

Note: Values in parentheses are standard error; *** indicates significance at the 1% level.

shows the heterogeneity results of geographical location. The direct effect of the interaction coefficient RCS × YJ is significantly negative, indicating that RCS has an obvious deterrent effect on GTI in riverine cities compared to non-riverine cities. Under RCS for managing the river basin water environment, riverine cities will face stronger environmental regulations than non-riverine cities, making them bear greater pressure for environmental management, and the increased cost of pollution control will create an “innovation crowding out effect”, thus inhibiting GTI.

4.7.4. Heterogeneity Analysis of the Watershed Basin

As one of the most developed and high-density economic corridors in the Yangtze River Basin, the YREB spans the east and west of China. Due to the unbalanced development in the upper, middle, and lower reaches of the YREB, the differences in economic development, geographical location, and resource endowment among different regions make each region implement differentiated governance strategies in the specific implementation process of RCS, which makes for different governance effects in different regions. Therefore, this paper divides YREB into the upper, middle, and lower reaches for difference analysis. The lower reaches include Shanghai, Jiangsu, Zhejiang, and Anhui provinces (cities). The middle reaches include Jiangxi, Hubei, and Hunan provinces (cities). The upper reaches include Chongqing, Sichuan, Guizhou, and Yunnan provinces (cities).

Table 10. Heterogeneity results of the watershed basin.

Variable	Upper Reaches		Middle Reaches		Lower Reaches
	Direct Effect	Indirect Effect	Direct Effect	Indirect Effect	Direct Effect	Indirect Effect
RCS	0.2035 *	0.4855 ***	0.3841 ***	0.1213	0.2491 ***	0.1062
	(0.1054)	(0.1527)	(0.0959)	(0.1968)	(0.0703)	(0.1684)
FD	0.0350	1.2306 ***	0.5955 ***	0.9509 ***	0.4666 ***	0.6114 ***
	(0.0787)	(0.1715)	(0.0877)	(0.2752)	(0.0704)	(0.1609)
IS	−0.0119	−0.8755 ***	−0.2849 **	−1.5310 ***	−0.3733 ***	−0.7176 **
	(0.1201)	(0.2187)	(0.1152)	(0.3214)	(0.0882)	(0.2861)
Open	3.3900	−25.3238 ***	−3.5538	5.0838	7.5271 ***	0.2280
	(3.4610)	(7.9540)	(2.6065)	(10.2700)	(1.2663)	(3.5106)
Edu	0.1620	1.2502 ***	0.2543 **	1.0427 **	0.8176 ***	0.7673 *
	(0.1030)	(0.2967)	(0.1076)	(0.4274)	(0.1167)	(0.4049)
Rod	0.0790 ***	0.1811 ***	0.0937 ***	0.3206 ***	0.1332 ***	0.1550 ***
	(0.0258)	(0.0507)	(0.0230)	(0.0961)	(0.0133)	(0.0498)
GS	23.3038 ***	46.0648 ***	4.7631 **	25.9835 ***	9.1870 ***	32.8256 ***
	(5.4079)	(12.8375)	(2.3115)	(7.4337)	(1.8997)	(4.2760)
ρ	0.3634 ***		0.5233 ***		0.5181 ***
	(0.0357)		(0.0411)		(0.0386)
σ2	0.2039 ***		0.1578 ***		0.1559 ***
	(0.0131)		(0.0095)		(0.0088)
City Effect	Yes		Yes		Yes
Year Effect	Yes		Yes		Yes
N	496		576		656
R2	0.8209		0.8687		0.9011
Log-like	−321.9056		−303.3737		−341.3586
AIC	703.8112		666.7474		742.7171
BIC	830.0085		797.4306		877.3019

Note: Values in parentheses are standard error; *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

shows the heterogeneity results of the watershed basin. The RCS can effectively stimulate the local GTI in the whole YREB. However, the effect of RCS on GTI is stronger in the middle and lower reaches. That is because the middle and lower reaches have a higher level of economic development and stronger economic strength to support enterprises to carry out GTI when implementing the RCS. The indirect effect of RCS on GTI is significantly positive only in the upper reaches, indicating that the policy effect of RCS on GTI has a positive spillover in the upper reaches.

4.8. Mechanism Test

The theoretical analysis shows that there are government governance effects, official incentive effects, and social supervision effects in the influence of RCS on GTI. In this section, we will empirically test the influence mechanism of RCS on GTI. The interaction terms of mechanism variables and core explanatory variables were added to model (2), and the following model (17) was constructed to prove the reliability of the mechanism.

$$\begin{array}{c}GT{I}_{it}={\beta }_0+\rho W\times GT{I}_{it}+{\beta }_{1}RC{S}_{it}+{\beta }_{2}RC{S}_{it}\times M_i+{\lambda }_i{X}_{it}+{\theta }_{1}W\times RC{S}_{it}\\ +{\theta }_{2}W\times RC{S}_{it}\times M_i+{\gamma }_{i}W\times X_{it}+{\mu }_i+{\nu }_t+{\epsilon }_{it}\end{array}$$

(17)

where M_i represents the mechanism variables, including government environmental governance (EG), official incentive (Off), and social supervision (Public). Other variables are the same as Model (2).

EG uses the proportion of environmental protection-related word frequency in the total word frequency of local government work reports as the proxy variable [48]. Off uses the official age as the proxy variable and is measured by dummy variables. Studies have shown that local officials are generally promoted at less than 55 years old [49]. Therefore, local officials under the age of 55 have stronger promotion incentives. So, we define Off as 1 if officials are younger than 55 years old, and 0 if otherwise. Public uses the Baidu environmental pollution search index as the proxy variable [50]. Due to data availability, the data range of Public is from 2011 to 2019, and municipalities are excluded.

Table 11. Mechanism test.

Variable	EG		Off		Public
	(1)	(2)	(3)	(4)	(5)	(6)
	Direct Effect	Indirect Effect	Direct Effect	Indirect Effect	Direct Effect	Indirect Effect
RCS × M	0.2377 **	−0.0679	0.0412 **	0.0057	0.0024 ***	−0.0006
	(0.0923)	(0.2517)	(0.0499)	(0.1670)	(0.0009)	(0.0021)
RCS	0.0397	0.4316 **	0.2252 ***	0.5831 ***	0.0971 **	0.2241 ***
	(0.0790)	(0.2032)	(0.0638)	(0.1708)	(0.0469)	(0.0870)
FD	0.2375 ***	0.8990 ***	0.2334 ***	1.1370 ***	0.1518 **	0.8237 ***
	(0.0445)	(0.1026)	(0.0446)	(0.1131)	(0.0596)	(0.1181)
IS	−0.2576 ***	−1.0227 ***	−0.2721 ***	−1.2824 ***	0.0634	0.3836 **
	(0.0595)	(0.1462)	(0.0600)	(0.1636)	(0.0804)	(0.1596)
Open	3.9453 ***	0.3984	3.9825 ***	3.8428	−1.1285	−18.4806 ***
	(1.0511)	(2.7223)	(1.0450)	(2.9580)	(1.7743)	(3.8756)
Edu	0.3525 ***	1.5023 ***	0.3473 ***	1.8554 ***	0.1970 **	0.9115 ***
	(0.0612)	(0.1910)	(0.0614)	(0.2214)	(0.0846)	(0.2146)
Rod	0.1098 ***	0.2040 ***	0.1085 ***	0.3162 ***	0.0608 ***	0.0682 **
	(0.0109)	(0.0308)	(0.0109)	(0.0357)	(0.0134)	(0.0317)
GS	7.9040 ***	28.7461 ***	7.7298 ***	36.3965 ***	2.1763	7.6582 **
	(1.2099)	(3.1491)	(1.2124)	(3.2528)	(1.3340)	(3.6697)
ρ	0.4501 ***		0.4486 ***		0.3527 ***
	(0.0207)		(0.0208)		(0.0332)
σ2	0.1757 ***		0.1765 ***		0.0946 ***
	(0.0061)		(0.0061)		(0.0044)
City Effect	Yes		Yes		Yes
Year Effect	Yes		Yes		Yes
N	1728		1728		954
R2	0.8672		0.8668		0.7439
Log-like	−997.7243		−1001.4646		−244.4173
AIC	2063.4487		2070.9292		556.8346
BIC	2248.9092		2256.3897		722.0972

Note: Values in parentheses are standard error; ** and *** indicate significance at the 5% and 1% levels, respectively.

reports the results of the mechanism test. According to the test results, the indirect effects of these mechanisms are not significant. Therefore, we analyze the mechanisms based on direct effects. Column 1 reports the influence mechanism of EG. The coefficient of RCS × EG is positive and statistically significant at the 1% level, indicating that EG can enhance the promoting effect of RCS on GTI. Column 3 reports the influence mechanism of Off. The coefficient of RCS × Off is significantly positive, indicating that with the enhancement of official incentives, the RCS has a stronger effect on promoting GTI. Column 5 reports the influence mechanism of Public. The results show that the improvement of social supervision intensity can enhance the effect of RCS on GTI. In conclusion, the three mechanisms of RCS affecting GTI are valid.

5. Conclusions and Policy Implications

Taking the RCS as a quasi-natural experiment, based on the panel data of 108 cities in the YREB from 2004 to 2019, this paper uses the progressive DID method and SDID method to verify the causal effect of RCS on GTI. The main conclusions are as follows:

(1) GTI in the YREB shows a rapid growth trend, and the lower reaches are generally higher than the middle and upper reaches. (2) The RCS has a significant positive incentive effect on GTI in both the region and its neighboring regions, and GTI itself can have a positive spillover effect on the neighboring regions. (3) The impact of RCS on GTI has significant policy characteristics and regional heterogeneity. Firstly, the spillover effect of RCS on GTI in the adjacent untreated area is stronger than that in the treated area. Secondly, compared with the forced vertical diffusion form, the spontaneous parallel diffusion form has a more obvious incentive effect on the local GTI, but it has an “innovation crowding out effect” on the GTI in its neighboring areas, which inhibits the GTI in its neighboring areas. Thirdly, the RCS hinders GTI in riverine cities. Fourth, in the whole YREB, RCS can effectively stimulate the local GTI, but only generate positive spatial spillovers in the upper reaches. (4) Government governance, official incentive and social supervision can enhance the promoting effect of RCS on GTI.

According to our conclusions, some useful policy implications are provided below:

(1)

With the implementation of RCS as the driving force, we should establish a GTI system, promote the transformation of environmental governance from terminal governance to whole-process prevention and control, and give full play to the positive role of technological innovation in environmental governance. The conclusion shows that RCS can promote GTI. Therefore, the policy effect of RCS should be given full play to promote regional GTI. On the one hand, we should improve RCS management, implement the regulations of RCS management, and strictly carry out pollution emission standards to force polluting enterprises to carry out GTI and use emission reduction technology to reduce pollution emissions. Innovatively implementing water rights trading should be adopted to guide and encourage enterprises to save water resources through green and energy-saving technologies. At the same time, through innovation subsidies, tax reduction and other supporting policies to improve the enthusiasm of enterprises for innovation, there should be moves to establish and improve the green financial system and an innovative product market trading system to create a good environment for green innovation, promote the green transformation of industries and the development of emerging green industries, as well as improve the level of regional green innovation. On the other hand, we should empower environmental governance with technology and actively promote the exploration of GTI in the fields of environmental quality monitoring and early warning systems, as well as ecological product value assessment, and to improve the efficiency of environmental governance and promote the development of modernization of environmental governance level and governance capacity.

(2)

All regions should improve the enthusiasm of environmental governance in, and promote the transformation of, environmental governance from passive to active governance. The results show that the regions with spontaneous RCS governance more actively promote regional GTI. Local governments of spontaneous regions usually have stronger environmental governance capabilities, and improvement in the government’s environmental governance level can enhance the promoting effect of RCS on GTI. Therefore, we should improve the environmental governance level of local governments to enhance the governance effect of the RCS. On the one hand, it is necessary to fully integrate command-and-control environmental regulation, market-incentive environmental regulation, and public participation environmental regulation, as well as to complete the environmental regulation system, increase the intensity of environmental regulations, guide the whole society to form a consensus on green development, and promote the development of energy conservation and emission reduction. On the other hand, we should strictly enforce the accountability and promotion incentive mechanism of RCS, improve the subjective initiative of RCS governance for local governments through multiple means, encourage enterprises to actively fulfill environmental protection responsibilities, enhance the momentum of GTI for enterprises, and promote the development of green innovation in the region.

(3)

We should formulate specific implementation plans for RCS that are suitable for the development of various regions according to local conditions. The unbalanced development of each region in the YREB makes the green incentive effect of RCS show obvious differences in different regions. Therefore, the normalized environmental governance pattern is not suitable for all regions. Each region should formulate environmental governance measures suitable for its own development according to its actual situation, so as to achieve hierarchical governance of environmental problems. At the same time, we should give full play to the knowledge spillover effect of RCS on the neighboring regions. Each region should actively learn the excellent experience of other cities in environmental governance and carry forward the spirit of “learning by doing” and explore the most suitable environmental governance mode for the development of its own region. The transboundary nature of rivers makes it inevitable that there will be problems of cross-regional coordination in the process of promoting the RCS, so it is important to advance the issue of inter-regional communication while strengthening intra-regional environmental management.

However, there are some limitations that require further research in the future. Firstly, our study is based on prefecture-level data. Different counties have different performances in the exploration practice of RCS. Future studies should be extended to the county-level to obtain more robust conclusions. Secondly, due to the lack of empirical data, this paper has not studied the impact of the implementation intensity of RCS on GTI. In a follow-up study, with the continuous accumulation of empirical data, the impact of the implementation intensity of RCS and regional policy implementation differences on GTI can be further studied.