Can Horizontal Ecological Compensation Improve the Water Environment in Cross-Provincial Watersheds?

Abstract

Horizontal ecological compensation is an important institutional innovation to promote ecological civilization and is the main functional area strategy in China under the current management mechanism. In this paper, we use contract theory to analyze the advantages of horizontal ecological compensation in cross-provincial watershed governance issues based on the basis of the case of Xin’anjiang River Basin, which is the first pilot horizontal ecological compensation in cross-provincial watersheds in China. We select panel data from 2004 to 2020 and use the synthetic control method (SCM) from the perspective of the water environment to assess the horizontal ecological compensation on the intensity of water pollution in the Xin’anjiang River basin. It is found that: (1) The horizontal ecological compensation can effectively respond to the failure of traditional ecological compensation methods in cross-provincial watershed governance, and, to some extent, solve the problem of lacking of incentives for upstream and downstream governments to participate; (2) The horizontal ecological compensation can reduce the water pollution intensity and improve the water environment in the Xin’anjiang River Basin; (3) However, the impact on the upstream is significantly lower than that of the downstream, and the policy effect on the upstream tends to be zero after the central government removes the subsidies to the upstream in 2018; and (4) The policy shows an expected effect two years ahead of the implementation. The findings of this paper have important implications for the promotion of the horizontal ecological compensation policy and the management of the water environment in cross-provincial watersheds.

1. Introduction

Over the last decades, the ecological damage and environmental pollution of river basins have caused great losses all over the world. Securing watershed ecosystem services is regarded as one of the fundamental challenges facing humanity [1]. Generally, one river basin may cover a wide area and span different administrative regions, and also involve many stakeholders [2,3]. Therefore, due to the varying economic interests, the watershed stakeholders each seek different ecosystem services and resources to ensure their individual socio-economic development goals, which will lead to unsustainable development of the watershed [4]. The emergence of watershed ecological compensation systems provides a solution to this problem [5,6]. Ecological compensation of river basins can be divided into two types: the “Pigouvian approach” based on the government intervention and the “Coasean approach” based on the market mechanisms [7,8]. The Coasean approach emphasizes the creation of a voluntary or market-based transaction for ecosystem services [9,10,11]. Alternatively, the Pigouvian approach allows for government intervention such as through regulation, tax, or subsidy [12,13].

However, in China, where water resources are owned by the State, there is a water resource management mechanism combining watershed management with administrative regional management. With regards to the management mechanism, the local government in charge of the management of the affected river basin is directly compensated by the central government. Due to the lack of effective regulatory and incentive mechanisms, the Pigouvian approach tends to result in problems with compensation in China; for example, local governments in upstream areas pursue compensation while local governments in downstream areas pursue “ride-sharing” [14].

The core of the Coasean approach is to internalize the externality of water resources products through clearly defining property rights and market transactions [15]. There are three problems when the Coasean approach is adopted in inter-provincial basin management [15]: firstly, given that the state owns the water resources in China, water users in the upstream and downstream areas have the right to use water resources. Through inter-provincial basin management, conflicts arise between upstream water users since they wish to promote development, while the environmental costs are borne by downstream water users [16]. As a result, there is a dilemma in defining initial property rights in terms of the Coasean approach [17]. Secondly, the Coasean ecological compensation relies on a principal-agent contract between the upstream and downstream water users, in which the central government relies on local government interventions, and downstream water users rely on upstream users to conserve and improve the water resources. However, the agreement of principal-agent contract must meet both the participation constraint and the incentive compatibility constraint (ICC), which means that the compensation standard should consider the opportunity cost of the upstream conservation efforts and the value of downstream water ecosystem services [18,19]. Thirdly, there will be information asymmetry, as it is difficult for downstream water users to quantify the opportunity cost of the upstream conservation efforts and for upstream users to quantify downstream ecosystem services; as such, a reasonable compensation standard cannot be determined [20]. Meanwhile, the downstream governments may fear exploitation by the upstream water users. Ultimately, this could lead to failure of negotiations of a principal-agent contract for ecological compensation. On this basis, there is a need to improve the existing model for basin ecological compensation and application in instances of state ownership of natural resources.

It is important to find suitable water environment management methods to adapt to China’s economic conditions and the administrative system [21]. To address devastating water environmental crises and to improve the quality of economic developments, China has already implemented multiple regional and national policies [22]. In 2012, China proposed a horizontal ecological compensation system that combines the Pigou approach and the Coasean approach. The first pilot was carried out in Xin’anjiang River Basin. The Horizontal Ecological Compensation pilot project in provincial and municipal governments across river basins is based on a two-way compensation model. If the upstream region provides the downstream with the required water quality, the downstream will provide corresponding compensation to the upstream; however, if the upstream region cannot meet the water quality requirements, the upstream will be asked to compensate the downstream. The horizontal ecological compensation is thus the integration of both the Pigouvian approach and the Coasean approach [23]. Governments of the upstream and the downstream can reach endogenous equilibrium in terms of water quality standards through negotiation. Therefore, the evaluation of the environmental effects of this new watershed ecological compensation system has become an issue worth studying.

In recent years, several studies on the evaluation of the effects of horizontal ecological compensation have been carried out. Zhang Jie (2018) conducted a theoretical analysis of horizontal ecological compensation in watersheds based on the contract theory and found that in the case of complete information, governments of the upstream and downstream can reach an endogenous equilibrium on water quality standards through negotiation, and in the case of incomplete information, government intervention is beneficial in reducing transaction costs [24]. Ke jiang, et al. proposed a tripartite evolutionary game model to study the evolutionary and steady-state strategies of watershed stakeholders [4]. Junichi Ito used a standard principal agent model to study the incentive design of the behavioral responses of ecosystem suppliers [25]. Yicheng Fu proposed a water resources allocation model and multi-objective optimization approach based on cooperative game (CG) level calculations [20]. In the literature on the effect of horizontal ecological compensation, Salzman points out that the horizontal ecological compensation can promote the rational utilization of the water resources, and reduce in consumption of water resources, improve the sustainable supply of ecosystem services, accelerate the participation of the upstream in ecological compensation and protection of the environment, and enhance the ecological and the social benefits of ecological compensation [26]. Meanwhile, Kosoy and Immerzeel et al. suggest that by providing incentives, the horizontal ecological compensation could encourage upstream water users to consider the downstream interests in decision-making, thereby reducing the consumption of water resources and improving the water quality in the river basin while balancing upstream and downstream interests [27,28].

Meanwhile, there are five common methods to evaluate the effects of watershed ecological compensation systems, such as the instrumental variable method (IV) [29], breakpoint regression (RD) [30], propensity score matching method (PSM) [31], data envelopment analysis (DEA) method [1,22] and difference-in-difference method (DID) [30,32,33]. However, these methods have the following drawbacks when applied to the assessment of horizontal ecological compensation policies in watersheds: the IV method is a very good way to deal with endogeneity, but it is difficult to find suitable instrumental variables; the RD method measures the local average effect around a critical value, not an overall average effect, and it is difficult to extend to overall studies. The DEA method requires a large sample size and is only valid while the sample is large enough. PSM and DID are widely used in policy evaluation [31], but they have a common difficulty in that they need to find a suitable control group, while it is difficult to find two similar watersheds in watershed horizontal ecological compensation assessment. Therefore, this paper uses contract theory as the theoretical basis and the synthetic control method as the empirical method to quantitatively assess the environmental effects of horizontal ecological compensation.

Based on the existing research, this paper constructs a comprehensive analysis framework based on the principal-agent contract and demonstrates the governance advantages of the horizontal ecological compensation in the basin, and quantitatively evaluates the environmental effects of horizontal ecological compensation by using the synthetic control method. Compared with previous literature, this paper is innovative in the following aspects: (1) This paper proposes that different ecological compensation models are essentially contracts. The horizontal ecological compensation model is analyzed in terms of contractual theory in order to highlight the advantages in terms of risk-taking and incentive compatibility; (2) Through adopting the latest Synthetic Control Method (SCM) in the evaluation of the effectiveness of the horizontal ecological compensation in the river basin, we overcame sample selection errors that may occur in the selection of the control group objects, in addition to endogeneity problems; and (3) We obtained some interesting conclusions, such as that the horizontal ecological compensation policy has a greater impact on the downstream than on the upstream, and the impact on the upstream disappears after the central government removes the subsidy.

The paper is organized as follows: Section 2 describes the theoretical analysis and model framework design; Section 3 provides the empirical design, and Section 4 presents the conclusions of this study.

2. Theoretical Framework

Ecological compensation is implemented at the level of provincial, municipal and county administrative units, and the upstream and downstream governments form a principal-agent relationship. For example, at provincial-level basin ecological compensation, the central government acts as the principal while the provincial government acts as the agent, and the former entrusts the latter to carry out ecological protection. The goal is to provide ecosystem services equal to or higher than the corresponding standards. The government’s goals for the ecological protection of the river basin are closely related to the incentive conditions in the contract. The provincial government needs to weigh the costs and benefits under different levels of efforts, and finally decides on the level of effort required for the ecological protection of the basin. This approach is applied at the provincial, municipal, and county level. Therefore, the ecological compensation mechanism of the river basin is based on the agency contract. In China, the agency contract focuses on the upstream and downstream governments. As such, this paper takes the principal-agent model as the basic theoretical framework and analyzes the effectiveness of the horizontal ecological compensation in terms of management of the water resources in inter-provincial basins.

2.1. Basic Hypotheses

Hypothesis 1 (H1).

It is assumed that the local government of the downstream, which acts as the principal, is risk-neutral, while the local government of the upstream, whicht acts as a proxy, is risk-averse; this is represented by the absolute risk aversion function $u=-e^{-\rho w}$, where ρ is the absolute risk aversion coefficient.

Hypothesis 2 (H2).

Asymmetric information between the upstream and downstream governments, that is, the behavior of the government of the upstream cannot be directly observed by the government of the downstream, and the linear contracts can only be provided on the basis of changes in water quality, $w_u=\alpha +\beta \pi$, where $w_u$ represents the total ecological compensation fund of the upstream local government, $\alpha (\alpha >0$) represents the fixed compensation fund provided by the central government, and $\beta$ represents the proportion of compensation funds obtained by the upstream government from the downstream government for the change of water quality in the section, and $\pi$ indicates the water quality of the section between the upstream and downstream governments.

Hypothesis 3 (H3).

Assume that A represents a combination of all the available interventions of the upstream government, and $a\in A$ represents a particular intervention. Specifically, in the ecological compensation of the river basin,$a$ represents the level of effort of the upstream government in the ecological protection of the river basin. The water quality of the basin section is $\pi =k_{1}a+\theta$, and the corresponding ecological value is $X=k_{2}a+\theta$, where $ka$ is the influence of the interventions of the upstream government on the water quality, and $\theta$ is an exogenous variable that indicates the influence of natural environmental conditions on the water quality of the basin which are not controlled by upstream and downstream governments; this gives the normal distribution of $\theta ~N\left(0,{\sigma }^2\right)$. The cost of the water quality control in the upstream government is $C=c{a}^2/2$. The upstream government has a reserved utility level $\overline{w}$, and when the upstream government attains expected utility below this level, it will refuse to participate in the ecological compensation program. The Parameter Description was list in

Table 1. Parameter Description.

Parameter	Description
$\alpha$	Fixed compensation funds provided by the central government for the upstream government
$\beta$	The proportion of compensation funds obtained by the upstream government from the downstream government for a change in water quality in the section
$\pi$	Water quality in the basin section
$a$	Efforts of the upstream government to control the water quality of the basins
$f_L$	Expected income of the central government
$\overline{w}$	Retention utility level of the local government
$L$	Random income of the central government
$w_u$	Random income of the upstream government $F_i$
$w_d$	Random income of the downstream government $F_i$
$f_u$	Expected income of the upstream government
$f_d$	Expected income of the downstream government
$\theta$	Other random factors affecting the water quality
$X$	Ecological value generated by the upstream government in governing the basin
$k_1, k_2$	Water quality control/ecological value effect coefficient of the upstream government
$c$	Cost coefficient of the water quality management of the upstream government
$\rho$	Risk aversion coefficient of the upstream government
$f_L$	Expected returns of the central government
$\mathrm{Var}\left(w_u\right)$	The risk that the upstream government assumes in collaborative decision-making
$\mathrm{Var}\left(X-w_u\right)$	The risk that the downstream government assumes in collaborative decision-making

as bellow.

2.2. Model Establishment

Based on the above hypothesis, the final utility function of the local upstream and downstream governments can be established. The random income of the upstream government $w_u$ is:

$$E\left(w_u\right)=E\left(\alpha +\beta \pi -1/2c{a}^2\right)=E\left(\alpha +\beta \left(k_{1}a+\theta \right)-1/2c{a}^2\right)$$

(1)

According to the deterministic equivalence income principle, the upstream government, which is risk-averse, has a utility expectation of $w_u$:

$$f_u=E\left(w_u\right)=\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2$$

(2)

The stochastic benefits of the downstream government are:

$$E\left(w_d\right)=E\left(k_{2}a+\theta -\alpha -\beta \pi \right)$$

(3)

As the downstream government is risk-neutral, its benefits are equal to the utility:

$$f_d=E\left(w_d\right)=-\alpha +\left(k_2-\beta k_1\right)a$$

(4)

The upstream and downstream local governments will pursue the maximization of their own utility, and the utility of the upstream government is not lower than the utility for its own use. The model is:

$$\max f_d=-\alpha +\left(k_2-\beta k_1\right)a$$

(5)

$$\mathrm{s}.\mathrm{t}.a\in argmax \left(\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\right);$$

(6)

$$\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\ge \overline{w}.$$

(7)

Formula (5) is the maximization of the income of the central government; Formula (6) is the ICC, that is, the local government determines the degree of efforts to control basin pollution on the basis of the proportion of the central government’s compensation funds to maximize its utility; Formula (7) is the participation constraint, that is, the effectiveness of the local government participating in the basin ecological compensation governance cannot be lower than its reserve utility, otherwise the local government will refuse participation in the basin ecological compensation (under the Chinese administrative system, the local government, as the lower level, generally will not directly refuse to participate in the ecological compensation proposed by the central government, but if the effectiveness of participating in the basin ecological compensation is lower than the effect of retaining its own use, the local government will refuse to participate in other indirect ways, which is equivalent to “direct rejection” in terms of the governance effect).

2.3. Model Discussion

This paper investigates whether the introduction of the horizontal ecological compensation has advantages in the regulation of the environment based on the state management mechanism of water resources, which combines basin management and administrative regional management and where there are no clear legal provisions in terms of the relationship between the property rights of the resource and of the environment. We analyzed two paradigms:

• Coase Paradigm Analysis of the Free Negotiation of “Upstream Government + Downstream Government”

When the upstream and downstream governments negotiate freely without the participation of the central government, $\alpha =0$, the ecological compensation fund obtained by the upstream government is: $w_u=\beta \pi$. The model can be simplified to:

$$\max f_d=\left(k_2-\beta k_1\right)a$$

(8)

$$\mathrm{s}.\mathrm{t}.a\in argmax \left(\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\right);$$

(9)

$$\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\ge \overline{w}.$$

(10)

Under the Kuhn-Tucker condition, the participation constraint can be derived from Formula (10). The ICC is derived from $\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2$ where $a$: $a=\beta k_1/c$. Substituting the constraints into the objective function, we get:

$$f_d=-k_1{}^2/c·{\beta }^2+k_1{k}_2/c·\beta$$

(11)

Letting the $d{f}_d/d\beta =0$, we obtain the optimal ecological value sharing of the upstream government $\beta$:

$${\mathsf{\beta }}_1=k_2/k_1$$

(12)

Then the risk that the upstream government assumes in collaborative decision-making is:

$$\mathrm{Var}\left(w_u\right)=\mathrm{Var}\left(\alpha +\beta \pi \right)={\beta }_1{}^2{\sigma }^2=k_2{}^2/k_1{}^2·{\sigma }^2$$

(13)

The risk that the downstream government assumes in collaborative decision-making is:

$$\mathrm{Var}\left(X-w_u\right)=1-\mathrm{Var}\left(w_u\right)=1-k_2{}^2/k_1{}^2·{\sigma }^2$$

(14)

In summary, under the situation of incomplete information and in the Coase paradigm of the free negotiation between the upstream and downstream governments, the optimal risk-bearing contract can be reached on the basis of the principal-agent contract. The downstream government should share the ecological value with the upstream government with the ratio ${\mathsf{\beta }}_1=k_2/k_1$. Under the condition of the optimal contract, the risk assumed by the upstream government will be $k_2{}^2/k_1{}^2·{\sigma }^2$.

2.

Analysis of the horizontal ecological compensation of “Central Government + Upstream and Downstream Government”

In the horizontal ecological compensation, the participation of the central government in the decision-making model is mainly through providing ecological compensation funds to the upstream governments. According to the current horizontal ecological compensation pilot in the basins of China, the central government will allocate the matching funds to the upstream governments for basin management regardless of water quality compliance in the section. Therefore, in the model, the compensation fund obtained by the upstream government is $w_u=\alpha +\beta \pi$. The model is:

$$\max f_d=-\alpha +\left(k_2-\beta k_1\right)a$$

(15)

$$\mathrm{s}.\mathrm{t}. a\in argmax \left(\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\right);$$

(16)

$$\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\ge \overline{w}.$$

(17)

Using the Kuhn-Tucker condition, the participation constraint is derived from Formula (17). The ICC is derived from $\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2$, where $a$: $a=\frac{\beta k_1}{c}$. Substituting the constraints into the objective function, we obtain:

$$f_d=-\frac{k_1{}^2}{2c}{\beta }^2-1/2\rho {\beta }^2{\sigma }^2+\frac{k_1{k}_2}{c}\beta -\overline{w}$$

(18)

Let $d{f}_d/d\beta =0$, we obtain the ecological value sharing of the upstream government $\beta$:

$${\mathsf{\beta }}_2=\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}$$

(19)

Then the risk that the downstram government assumes in collaborative decision-making is:

$$\mathrm{Var}\left(w_u\right)=\mathrm{Var}\left(\alpha +\beta \pi \right)={\mathsf{\beta }}_2{}^2{\sigma }^2={\left(\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}\right)}^2{\sigma }^2$$

(20)

The risk that the downstream government assumes in collaborative decision-making is:

$$\mathrm{Var}\left(X-w_u\right)=1-\mathrm{Var}\left(w_u\right)=1-{\left(\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}\right)}^2{\sigma }^2$$

(21)

In summary, in the horizontal ecological compensation under the situation of incomplete information, the optimal risk-bearing contract between the central government and the upstream/downstream government can be reached on the basis of the principal-agent contract. The downstream government should share the ecological value to the upstream government with the ratio ${\mathsf{\beta }}_2=\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}$. Under the condition of the optimal contract, the risk assumed by the upstream government is ${\left(\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}\right)}^2{\sigma }^2$.

2.4. Model Conclusion

1.

The participation of the central government reduces the risks of the upstream and downstream local governments.

When the upstream and downstream governments negotiate freely, the downstream government should share the ecological value to the upstream government with the proportion of ${\mathsf{\beta }}_1=\frac{k_2}{k_1}$; when the central government joins and implements the horizontal ecological compensation mechanism, the downstream government should share the ecological value to the upstream government with the proportion of ${\beta }_2=\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}$. As the $k_1,k_2,c,\rho and {\sigma }^2$ are all greater than 0, ${\beta }_2=\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}<\frac{k_1{k}_2}{k_1{}^2}=\frac{k_2}{k_1}={\beta }_1$, that is ${\mathsf{\beta }}_2<{\mathsf{\beta }}_1$, and ${\left(\frac{k_1{k}_2}{k_1{}^2+c\rho {\sigma }^2}\right)}^2{\sigma }^2<\frac{k_2{}^2}{k_1{}^2}{\sigma }^2$. This indicates that, in the horizontal ecological compensation of the river basin, the supporting funds of the central government not only reduced the proportion of the ecological value of the river basin governance that the downstream government needs to distribute to the upstream governments but also reduced the risks of the upstream and downstream governments.

2.

The matching funds provided by the central government support the incentive compatibility mechanism in the agency contract.

When the upstream and downstream governments negotiate freely, the upstream government obtains the benefits $f_u=\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2$, and the ICC reached by its entrusted agency contract is $\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\ge \overline{w}$. When the central government joins and implements the horizontal ecological compensation mechanism, the upstream government obtains the benefits $f_u=\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2$, and the ICC reached by its entrusted agency contract is $\alpha +\beta k_{1}a-1/2c{a}^2-1/2\rho {\beta }^2{\sigma }^2\ge \overline{w}$. For the $\alpha >0$, under the horizontal ecological compensation mechanism of the river basin, the benefits of the upstream government are more likely to be higher than its retained utility level $\overline{w}$, thus facilitating the agency contract and river basin governance.

Based on the results of the theoretical model analysis above, this paper proposes the theoretical proposition to be verified: the horizontal ecological compensation supports the management of water resources, where the state has a management mechanism for water resources that combines basin management and administrative regional management and where there are no clear legal provisions in terms of the relationship between the property rights of the resource and the environment.

3. Method

3.1. Synthetic Control Methods

The key to scientifically assessing the impact of the horizontal ecological compensation mechanism in the basin is to find a suitable assessment method for policy effects. In this paper, we selected the SCM proposed by Abadie et al. [34,35] to evaluate the effectiveness of the horizontal ecological compensation in China. The method defines the “counterfactual” control group of each policy intervention through the weighted average of the control group. The control object, the environment of the horizontal ecological compensation pilot without implementing the horizontal ecological compensation mechanism, is simulated to compare the implementation effectiveness. This is equivalent to a quasi-experimental study, in which the comparative test is conducted on the same basin at the same time during the study period. A comparison of the basin environment with and without the implementation of the horizontal ecological compensation show the effectiveness of the horizontal ecological compensation in terms of the ecological management of the river basin.

Assume that the Xin’anjiang River Basin covers N + 1 cities in the Zhejiang and Anhui Provinces. City 1 began to implement the horizontal ecological compensation in period $T_0$, while the other N cities did not implement the horizontal ecological compensation. $Y_{1it}$ indicates the ecological compensation in region $i$ during the period $t$. $Y_{0it}$ indicates the result where ecological compensation was not implemented in region $i$ during the period $t$, so the effect of the ecological compensation in region $i$ is ${\tau }_{it}=Y_{1it}-Y_{0it}$, where $i=1,\dots \dots ,\mathrm{N}+1,\mathrm{t}=1,\dots \dots T$. The result of the implementation of the horizontal ecological compensation in region $i$ during the period $t$ is $Y_{it}=D_{it}{Y}_{1it}+\left(1-D_{it}\right)Y_{0it}=Y_{0it}+{\tau }_{it}{D}_{it}$, $D_{it}$. If the horizontal ecological compensation has been implemented, the value is 1; otherwise, the value is 0. For convenience, it is assumed that the first region implements the horizontal ecological compensation in $T_0$ while other N regions did not implement in all periods, then for $t>T_0$, the effect of the horizontal ecological compensation can be expressed as ${\tau }_{1t}=Y_{11t}-Y_{01t}$. Since Region 1 implemented the horizontal ecological compensation, the potential result $Y_{11t}$ can be observed in the period of $t>T_0$, but the potential result $Y_{01t}$ cannot be observed where the horizontal ecological compensation is not implemented. To estimate the counterfactual for Region 1, $Y_{01t}$ can be represented by the following model (Abadie et al., 2010) [35]:

$$Y_{0it}={\delta }_t+{\theta }_t{Z}_i+{\lambda }_t{\mu }_i+{\epsilon }_{it}$$

(22)

where, ${\delta }_t$ is a time-fixed effect; $Z_i$ is an observable $\left(K\times 1\right)$ dimensional covariate, indicating a control variable that has not affected by the horizontal ecological compensation pilot; ${\theta }_t$ is a $\left(1\times K\right)$-dimensional unknown parameter vector, ${\lambda }_t$ is a common factor vector that cannot be observed in a $\left(1\times F\right)$ dimension, ${\mu }_i$ is a $\left(F\times 1\right)$ dimensional coefficient vector, and ${\epsilon }_{it}$ is a short-term shock that cannot be observed in each region with the assumption that the mean value is 0 at the regional level.

To solve $Y_{0it}$, an $\left(N\times 1\right)$ dimension weight vector $W=\left(w_2,\cdots ,w_{N+1}\right)$ can be considered to satisfy $w_j\ge 0,j=2,\cdots ,N+1$ and $w_2+\cdots +w_{N+1}=1$. Each specific value of the vector $W$ represents a composite control for Region 1, which is the weighted average of all regions within the reference group.

Abadie et al. (2010) [35] proved that for $T_0<t\le T$, the counterfactual of Region 1, as a result, the estimated effect of the horizontal ecological compensation can be represented by the synthetic control group i.e., $\widehat{Y_{01t}}={{\displaystyle \sum }}_{j=2}^{N+1}{w}_j^{\ast }{Y}_{jt}$, is:

$${\widehat{\tau }}_{1t}=Y_{1t}-{{\displaystyle \sum }}_{j=2}^{N+1}{w}_j^{\ast }{Y}_{jt},t\in \left[T_0+1,\dots T\right]$$

(23)

In order to minimize the distance $\left|X_1-X_{0}W\right|$ between $X_1$ and $X_{0}W$ to determine the weight $W^{\ast },$ the expression is $X_1-X_{0}W=\sqrt{{\left(X_1-X_{0}W\right)}^{’}V\left(X_1-X_{0}W\right)}$. $X_1$ is the $\left(m\times 1\right)$ dimensional eigenvector of the area before the implementation of the horizontal ecological compensation pilot; $X_0$ is the $\left(m\times N\right)$ dimensional matrix, and the column $j$ of $X_0$ is the corresponding eigenvector of the region $j$ before the implementation of the ecological compensation pilot. V is a $\left(m\times m\right)$ symmetric semi-definite matrix. Here, the method developed by Abadie et al. (2010) [35] is used to obtain $V$, so that the ecological environment of the synthetic area will approximate the pilot area before the implementation of the horizontal ecological compensation. The horizontal ecological compensation effect of the synthetic regions, obtained by weighting, simulates the ecological environment of the basin when where the horizontal ecological compensation is not implemented. The difference in the ecological environment of the river basin between the policy implementation area and the synthetic area is the impact of the horizontal ecological compensation on the ecological environment of the basin.

3.2. Overview of the Study Area

The study area is the Xin’anjiang River Basin, which is the first horizontal ecological compensation pilot in China. The Xin’anjiang River feeds into the Qiantangjiang River. The length of the main stream is 373 km and the drainage area is more than 11,000 km². The Xin’anjiang River spans two administrative regions, flowing from Anhui Province into Zhejiang Province and discharges into the Qiandao Lake. The length of the river in Hangzhou is 128 km, and in Huangshan City is 243.3 km. Most ecological impacts directly affecting the Xin’anjiang River Basin originate in Huangshan City in Anhui Province. For this reason, Huangshan City has closed sewage companies and lost many development opportunities. The amount of water flowing into the Zhejiang Province from the Anhui Province region accounts for 68% of the total water flowing into the Qiandao Lake, which has a significant impact on the water quality in Zhejiang Province. The ecological and economic relationship between the two Provinces is very clear. Economic development and associated funds for ecological compensation are greater in Zhejiang Province in the downstream areas. In 2012, under the guidance of the relevant national ministries and commissions, the Anhui and Zhejiang Provinces implemented the first horizontal ecological compensation pilot in the Xin’anjiang River Basin, comprising the upstream areas of Huangshan City and Jixi County in Xuancheng, and downstream areas of Chun’an County in Hangzhou.

To date, the horizontal ecological compensation pilot of the Xin’anjiang River Basin has been implemented over three three-year periods. The initial pilot period was 2012 to 2014. According to the “Xin’anjiang River Basin Water Environment Compensation Pilot Implementation Plan” of 2011, the compensation funds amount to 500 million yuan per year, of which the central government invests 300 million yuan, and the Anhui and Zhejiang provinces invest 100 million yuan, respectively. The compensation principle is as follows: the cross-border section of Anhui and Zhejiang provinces was used as the assessment monitoring section. If the water quality of the section is better than the basic limit, then 100 million yuan will be allocated to Anhui Province from Zhejiang Province, and if the water quality of the section is inferior to the basic limit or a major water pollution accident has occurred in the Anhui section of the Xin’anjiang River Basin, then 100 million yuan will be allocated to Zhejiang Province from Anhui Province. Regardless of the circumstances, the funds of the central government are allocated to Anhui Province. The basic water quality limit of the section is the average water quality of Xin’anjiang River in the first three years of the pilot. The water quality of the section is jointly monitored by the Environmental Protection Department of the two Provinces overseen by the national environmental monitoring department.

The second pilot study ran from 2015 to 2017. On the basis of the first round of pilot programs, the “double improvement” policy was implemented where both the funding subsidy standards and water quality assessment standards were raised. In terms of the compensation funds, the accumulated funds for the three years were 2.1 billion yuan, to which the central government contributed a total of 900 million yuan, through annual subsidies of 400 million yuan, 300 million yuan, and 200 million yuan. The Anhui and Zhejiang Provinces contribute 200 million yuan each year, respectively. The compensation index is measured by the water quality of the intersection, and the benchmark limit is adjusted to the average of the three-year joint monitoring results from 2012–2014. Incremental subsidies and the principle of “the better the water quality, the higher the price” are implemented.

The third round of pilot study is from 2018 to 2020. Compared with the implementation plans of the first two rounds of pilot projects, there are two major changes in the implementation plan of the new signing agreement: first, the water quality assessment standard is higher; and the water quality stability coefficient is improved from 89% of the second round to 90%; the weight coefficients of the four indicators of NH₃-N, chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) have adjusted from 25% to 22%, 22% and 28%, respectively. Secondly, the central government will no longer provide financial compensation. The Anhui and Zhejiang Provinces will contribute 200 million yuan each year, and the scope for use of compensation funds has been expanded. The compensation funds are used for comprehensive environmental management, water pollution control, ecological protection and construction, industrial restructuring, industrial layout optimization, and ecological compensation. Various incentives, such as the establishment of green funds, public-private partnership (PPP) models, and financing discounts have been implemented to improve the comprehensive management of the Xin’anjiang River Basin and the investment in green industries. In addition, special emphasis has been placed on reducing the runoff of chemicals from agricultural lands.

After two rounds of piloting, the effectiveness of the ecological compensation mechanism in the Xin’anjiang River Basin is evident: the overall water quality of the basin is excellent and stable, and the water quality of the Qiandao Lake has simultaneously improved. According to the Ministry of Ecology and Environment Planning Institute, the value of the ecosystem services in the Xin’anjiang River in 2008 was 24.65 billion yuan, and the value of the water ecosystem services was 6.45 billion yuan. The awareness of ecological and environmental protection amongst residents in the Basin has been enhanced, with an overall improvement in the quality of life.

3.3. Data

Hangzhou City, Huangshan City, and the Jixi County of Xuancheng City participated in the horizontal ecological compensation pilot project in the Xin’anjiang River Basin. In order to ensure the consistency of data, Jixi County was not considered, and Huangshan City and Hangzhou City were used as experimental groups. The assessment of the SCM requires that the ecological environment characteristics of other sample cities are similar to those of Huangshan City and Hangzhou City as far as possible and are not affected by the horizontal ecological compensation. Therefore, 25 other cities in Anhui Province and Zhejiang Province were selected as the reference group, namely Ningbo, Jiaxing, Huzhou, Shaoxing, Zhoushan, Wenzhou, Jinhua, Quzhou, Taizhou, Lishui, Hefei, Huaibei, Bozhou, Suzhou, Bengbu, Fuyang, Huainan, Chuzhou, Lu’an, Ma’anshan, Wuhu, Xuancheng, Tongling, Chizhou, and Anqing.

The water pollution intensity was selected as the explanatory variable to determine the effectiveness of ecological compensation [36]. The goal of ecological compensation in Xin’anjiang is to improve the water quality of the River Basin. The water pollution intensity was characterized by industrial wastewater discharge/real gross domestic product (GDP). There are two reasons for this: (a) at present, the industrial, domestic, and agricultural non-point source pollution are the main sources of river pollution in China. Industrial pollution, as the most important pollution source, originates mainly from industrial wastewater discharge. Therefore, industrial wastewater discharge is used to represent the river pollution. (b) Economic development and environmental protection are the dual goals of China at this stage.

In order to consider the fitting effect of the synthetic control object and the robustness of the results, the infrastructure (per capita road area), industrial structure (added value of the tertiary industry/real GDP increase), the level of openness (the actual use of foreign direct investment at the end of the year), the per capita GDP, and the water endowment (the total amount of water resources per capita in each area) were selected as predictive control variables. The choice of control variables follows three principles: (a) the control variables must be applicable to all experimental and reference groups, (b) the control variables should consider the root causes of changes in water pollution intensity, such as the water endowment and the industrial structure, and (c) the control variables should not be affected by the horizontal ecological compensation. Therefore, we include indicators such as infrastructure, openness and per capita GDP, as these indicators are closely related to economic development.

The above variable data are mainly derived from the “China Statistical Yearbook (2005–2021)”,” China City Statistical Yearbook (2005–2021)”, “Anhui Statistical Yearbook (2005–2021)”, “Zhejiang Statistical Yearbook (2005–2021)”, and “Statistical Bulletin”.

3.4. Results and Discussion

Figure 1 shows the average water pollution intensity of the experimental group (Huangshan and Hangzhou) and the control group from 2004 to 2020. It is evident that: (a) in 2012, before the implementation of the horizontal ecological compensation pilot, the water pollution intensity in Huangshan City was lower than the average water pollution intensity in Anhui Province and other cities in Zhejiang Province (except Huangshan City and Hangzhou City). However, the water pollution intensity in Hangzhou City was higher than the average water pollution intensity in other cities. (b) From 2010 to 2012, there was a downward trend in water pollution intensity in the experimental group (Huangshan and Hangzhou). (c) If the implementation period of the horizontal ecological compensation is set from 2010 to 2012, a significant decrease in water pollution intensity was observed in Huangshan and Hangzhou after the implementation of the horizontal ecological compensation when compared with the control group.

Given that the impact of the horizontal ecological compensation varies between upstream and downstream cities in the Xin’anjiang River basin, results are reported as Figure 2a the overall impact of the horizontal ecological compensation on the Xin’anjiang River Basin. Figure 2b the relative impact of the horizontal ecological compensation on upstream Huangshan and downstream Hangzhou.

3.4.1. Overall Impact of the Horizontal Ecological Compensation on the Xin’anjiang River Basin

Taking the length of the river section as the ratio of Huangshan and Hangzhou, a new city is synthesized on the Xin’anjiang River. Figure 3 shows the comparison of water pollution intensity between Xin’anjiang and synthetic Xin’anjiang before and after the implementation of the horizontal ecological compensation. The results are as follows: (a) there is a declining trend in water pollution intensity before the implementation of the horizontal ecological compensation, in both the actual and synthetic Xin’anjiang River Basin. Hence, without the implementation of the horizontal ecological compensation, the water pollution intensity was already showing a downward trend, largely due to the increasing awareness of environmental protection among the Chinese people in general, especially the series of environmental protection measures taken by the Chinese government. (b) However, in 2012, when horizontal ecological compensation was introduced, the two lines of the Xin’anjiang River Basin and its synthetic counterpart departed, and the water pollution intensity of the actual Xin’anjiang River Basin was lower than the synthetic Xin’anjiang River Basin. This means that the implementation of the horizontal ecological compensation had an impact on the reduction of water pollution intensity in the Xin’anjiang River Basin. (c) Another interesting finding in Figure 3 is that, in 2010, two years before the implementation of the horizontal ecological compensation, the water pollution intensity of actual Xin’anjiang was significantly lower than that of synthetic Xin’anjiang. It means that there is an expected effect in the implementation of the horizontal ecological compensation. This may be due to the extensive preliminary work done by the central government, Zhejiang Province and Anhui Province between 2010 and 2012. In March 2010, the Ministry of Finance and the Ministry of Environmental Protection (MOEP) consulted with Anhui and Zhejiang provinces on the implementation plan for the horizontal ecological compensation pilot project in the Xin’anjiagn River Basin, and allocated start-up funds of 50 million RMB to Huangshan City and Jixi County in December of the same year.

3.4.2. Upstream and Downstream Impacts of the Horizontal Ecological Compensation in Xin’anjiang River Basin

The effect of the horizontal ecological compensation varies between the upstream and downstream areas. Therefore, the impact of the horizontal ecological compensation is assessed separately for Huangshan City and Hangzhou City.

Figure 4 shows the effect of the horizontal ecological compensation on the water pollution intensity upstream and downstream of the Xin’anjiang River Basin. The water pollution intensity is set as a predictor variable and the vertical dashed line indicates the time when the horizontal ecological compensation was implemented.

Figure 4a shows that, before the implementation of the horizontal ecological compensation in 2012, the water pollution intensity of Huangshan resembled that of synthetic Huangshan, indicating that synthetic Huangshan is well fitted to the variation path of the water pollution intensity of Huangshan. After the implementation of the horizontal ecological compensation pilot in 2012, the two gradually deviated. The water pollution intensity of Huangshan is significantly lower than that of synthetic Huangshan. The difference between them is the effect of the horizontal ecological compensation pilot on water pollution intensity. This indicates that the implementation of the horizontal ecological compensation has significantly reduced water pollution in Huangshan.

The impact of the horizontal ecological compensation mechanism on the downstream Hangzhou City is shown in Figure 4b. Before the horizontal ecological compensation pilot was introduced, the water pollution intensity of synthetic Hangzhou resembled that of Hangzhou, indicating that synthetic Hangzhou is well fitted to the variation path of the water pollution intensity of Hangzhou. After 2012, the two gradually deviated. The water pollution intensity of Hangzhou is significantly lower than that of synthetic Hangzhou and maintained a downward trend, proving that the horizontal ecological compensation pilot significantly reduced the intensity of the water pollution in Hangzhou.

Meanwhile, Figure 4 shows that, between 2010 to 2012, there is a “GAP” between the actual water pollution intensity of Huangzhou and that of its synthetic counterpart, and the same phenomenon occurs in Huangshan, which indicates the expected effect. Therefore, in order to assess the effect of the horizontal ecological compensation policy more accurately, the time window was adjusted to 2010 and re-fitted. The results show in Figure 5, that the “GAP” disappeared. This result has two implications; firstly, it shows that the results in Figure 4 are robust; secondly, it shows that the time when the real effect of horizontal ecological compensation appears is 2010.

Table 2. Weight of Each City in Synthesized Huangshan.

City	Weight
Lishui	0.245
Chuzhou	0.090
Zhoushan	0.665
Else	0

and

Table 3. Weight of Each City in Synthesized Hangzhou.

City	Weight
Lishui	0.086
Chuzhou	0.160
Shaoxing	0.398
Zhoushan	0.078
Quzhou	0.278
Else	0

shows the weights of the synthetic control groups corresponding to Huangshan City and Hangzhou City. The results show that synthetic Huangshan is synthesized by Lishui, Chuzhou, and Zhoushan; Zhoushan has the highest weight of 0.665. Synthetic Hangzhou is composed of Lishui, Chuzhou, Shaoxing, Zhoushan and Quzhou; Shaoxing has the highest weight at 0398.

Based on the overall results, we can conclude that: (a) the implementation of the horizontal ecological compensation has the expected effect, and the policy took effect in the year of signing the agreement; (b) horizontal ecological compensation can significantly reduce the intensity of water pollution in the Xin’anjiang River basin, and the impact on the upstream is lower than that on the downstream; (c) Furthermore, the impact on the upstream first increased and then decreased, and approached 0 in 2018. However, the downstream effect changed little over time, with the mean remaining around −5.93.

3.5. Robustness Checks

Two robustness checks were conducted to test the reliability of the empirical results and prove that the water pollution differences are caused by the horizontal ecological compensation pilot instead of the other external factors.

3.5.1. Replacing the Pilot Implementation Year

In order to remove the effect of the policy implementation time on the empirical results presented in the previous section [37], this article replaces the policy implementation time from 2010 to 2011 for the robustness testing by the SCM method. Figure 6 shows that the fitting results are generally consistent with the previous paper, indicating that the previous results regarding the policy effect are robust independent of the year of policy implementation.

3.5.2. Placebo Test

In order to test the robustness of the conclusion in the previous section, a placebo test was conducted, as proposed by Abadie et al. (2010) [35].

The test was designed on the hypothesis that in 2010 the horizontal ecological compensation pilot was evident in all reference group cities and synthetic control objects were constructed for corresponding cities using the SCM. Against the backdrop of the expected effect in the hypothesis, the real effect of the horizontal ecological compensation in Hangzhou and Huangshang and the expected effect in all reference group cities from the hypothesis were compared. If there was a disparity between the two, the effect of the horizontal ecological compensation would be evident.

With a similar test method, we performed a placebo test in the other 25 cities. According to Liu and Fan (2021), a city with similar or different socio-economic characteristics can be chosen as the hypothetical treatment unit in a placebo test [38]. Therefore, we selected Zhoushan and Shaoxing as hypothetical treatment units because they had the biggest similarity weights with Huangshan and Hangzhou (see Table 2 and Table 3), and the cities Anqin and Huzhou that had a zero similarity weight with Huangshan and Hangzhou. The results of the placebo test are visualized in Figure 6.

As Figure 7 shows, the four cities display the same downward trend in water pollution intensity as Huangshan and Hangzhou over the period 2004–2020. And in the four cities, their actual water pollution intensity dips clearly and consistently below their synthetic water pollution intensity from 2010, with a few exceptions in specific years. However, there is no clear difference between their actual water pollution intensity and synthetic water pollution intensity, as the solid and dotted lines largely coincide. Based on these results, we conclude that the reduction of water pollution intensity in the Xin’anjiang River Basin was caused by the implementation of the horizontal ecological compensation in 2010.

4. Conclusions

Horizontal ecological compensation is an innovative means of watershed management in China, which is implemented in conjunction with river basin management and administrative regional management. This study evaluated the effectiveness of horizontal ecological compensation by the contract theory and the SCM method. The conclusions are as follows.

Theoretically, the horizontal ecological compensation is an effective watershed governance method in cross-provincial watersheds in China. Where the interplay between upstream and downstream governments does not meet the participation constraints and ICCs, the central government contributes supporting funds to address the “bid/ask” differences of the two sides and thus overcomes the problem of insufficient incentives in the Pigouvian model. The two-way compensation with both reward and punishment can solve the problem of the distribution of initial property rights, which helps to alleviate the limitations of the Coase model and to allow for effective management of the watershed.

The empirical results show that: (a) the horizontal ecological compensation can reduce the water pollution intensity in the Xin’anjiang River Basin and have a positive impact on the water environment; (b)the effect on the upstream environment is lower than the downstream. Specifically, the impact on the upstream first increased and then decreased, and approached 0 in 2018. However, the downstream effect changed little over time. This may be related to the fact that the central government no longer provides subsidies to the upstream after 2018; (c) In 2010, two years ahead the implementation of the horizontal ecological compensation pilot, there were differences in the fitting results of Huangshan and Hangzhou. This gap indicates the expectation effect. This may be due to the fact that the Ministry of Finance and the Ministry of Environmental Protection (MOEP) consulted with the Anhui and Zhejiang provinces on the implementation plan for the horizontal ecological compensation pilot project in the Xin’anjiagn River Basin in 2010. Through the effectiveness test, the robustness and reliability of the above research conclusions are further proved.