Mandatory Targets and Environmental Performance: An Analysis Based on Regression Discontinuity Design

Abstract

It is a critical question for environmental governance to examine whether the administrative award and punishment measures are effective in promoting environmental governance performance. Choosing the implementation of a mandatory target system (MTS) as the subject, this paper employs a fixed-effect panel data model and regression discontinuity design to test whether the MTS has improved the environmental governance performance of local governments in China. The results of this research demonstrate that the MTS has a positive effect on environmental performance, however the regression discontinuity design illustrates that the reward and punishment measures in the MTS have no significant effects on the provincial environmental performance. The results of this research provide a reasonable explanation to the existing gaps among the studies on the effectiveness of the MTS. This study has profound policy implications for the design and implementation of the environmental governance system in China.

1. Introduction

The mandatory target system (MTS) implemented during the 11th and 12th Five Year Plan (FYP) (2006–2015) is one of the most important mechanisms in the practice of environmental governance in China. The MTS adopted the method of multi-layer decomposition, in which the environmental objectives on unit energy consumption and major pollutant emissions are assigned to local governments, and the corresponding reward and punishment measures for the completion of objectives are stipulated in detail. This environment “Target Responsibility System” with Chinese characteristics has received great attention from both scholars and government officials since its implementation [1]. The environmental governance performance has improved remarkably, since the initiation and implementation of the MTS. During the 11th FYP (2006–2010), the average annual reduction in the energy consumption per unit of GDP, the total emission of sulfur dioxide (SO₂) and the total Chemical Oxygen Demand (COD) were 8.63%, 3.04% and 2.62% respectively. During the 12th FYP (2011–2015), the numbers for the three indicators were 6.28%, 2.50% and 2.62% respectively, all of which have achieved the expected progress. These results presented sharp contrasts to the dramatic increase in the three indicators during the 10th FYP (2001–2005).

The academic discussions on the MTS are mainly focused on two aspects. At the macro level, the relationship between the MTS and environmental performance improvement has been empirically examined through quantitative regression methods, typically by panel data analysis [2]. The results from this stream of research demonstrated that the MTS has a positive effect on the improvement of environmental performance [3,4,5,6]. At the micro level, methods such as field observations are generally adopted in the impact evaluation of the MTS. Contrary to the findings from macro level studies, the results from micro level analyses illustrated that in most cases, the reward and punishment measures of the MTS can be effectively evaded by local governments and their officials [7,8,9,10,11]. The conflicts between these findings demonstrate that further studies are needed to scrutinize the functions and outcomes of the MTS.

To fill this lacuna, this paper mainly focuses on addressing the following two questions. The first discusses whether the setting of mandatory targets has led to an enhancement in China’s environmental performance. This leads to Hypothesis I: the establishment of the MTS has a positive effect on provincial environmental performance. A fixed effect panel data model is utilized to test the influence of mandatory targets on each province’s energy consumption per unit GDP and the growth rate of COD and SO₂ emissions. The second question involves whether the impact of the MTS on environmental performance is driven by the rewards and punishment system, thus leading to Hypothesis II: if the rewards and punishment measures have an impact on local governments’ environmental performance, then the energy-saving and emission-reduction condition will be worse after the target completion reaches 100%. We then investigate the question by designing and implementing a Regression Discontinuity Design (RDD).

Results from this study showed that the setting of mandatory targets significantly promotes China’s environmental governance performance by reducing the provinces’ energy consumption per unit of GDP and the growth rate of COD and SO₂ emissions. With no significant decline of energy-saving and emission-reduction strength in local policies after full completion of the MTS target, we can conclude that the mandatory targets do not work through the implementation of formal rewards and punishment measures. In other words, gaining corresponding rewards or avoiding punishment is not the local governments’ major motivation to make them implement the energy-saving and emission-reduction measures to meet the targets.

The results of this research bridge the findings of macro level and micro level research on Chinese environmental policy, responding to the debate over the effectiveness of mandatory targets in previous literature. In the meantime, the results help us reinvestigate the relationship between policy implementation and performance management in China’s bureaucratic system, and contribute to our re-examination of the power relations and incentive mechanisms in authoritarian states.

The rest of the paper is organized as follows: we first offer a review of the literature on the relationship between the mandatory target and environmental performance. We then discuss the research design by presenting the empirical models and data. The results of the empirical analysis are presented and discussed next. The paper concludes with research findings and policy implications.

2. Literature Review

The study on how policy and institutional designs from the national government affect the implementation of environmental policies has been a central issue for policy process scholars [12,13,14]. In the context of Western Countries, the focus on the implementation effectiveness of centrally designed policies and goals is typically called a top-down approach to policy implementation [12]. Within this approach, it is typically argued that characteristics of the policy statutes, such as the clarity of policy directives and adequacy of causal theory, are related to effective policy implementation. Such top-down perspectives are also reflected in the early regulatory enforcement literature in public administration, which argued that the effectiveness of regulatory enforcement comes from a strong deterrence strategy embedded in the policy statutes [15]. What was missing in this literature is the motivations and behaviors of the local agents for implementation. The second wave of policy implementation scholars have thus focused on the bottom-up strategies in policy implementation, with explicit focus on the motivations of the street-level bureaucrats [16,17,18]. The development and application of performance management across government entities represents efforts to integrate the top-down approach with the bottom up approach to policy implementation [17]. It works in a way that performance evaluations of the local officials will be factored into the implementation process. While the literature discusses the performance management and policy implementation in general, little extant literature has examined such a relationship in the context of environmental policy. This study seeks to contribute to the above general policy of study literature by examining the role of performance management in the implementation of a centrally designed environmental policy in China.

In fact, the adoption of mandatory targets in environmental governance is not limited to China, but it is implemented on a global scale [19]. In the late 20th century and early 21st century, countries including the U.S. have extensively adopted the Renewable Portfolio Standards (RPS) or Renewable Obligation (RO) to raise their deployment level of renewables [20,21]. The governments tried to encourage more use of renewables in energy companies by granting certificates for every electricity unit produced by certified renewable energy generators. Despite its extensive adoption, there have been debates over its effectiveness. The study by Menz et al. pointed out that the RPS plays a positive role in the development of wind power [22], while Delmas and Wiser argued that the success of the RPS implementation relies heavily on its policy context and design details [23,24]. Furthermore, the study by Carley contended that, although the RPS could lead to an increase of renewable energy generation, it is not a significant predictor of the percentage of renewable energy generation out of the total generation mix [25]. Bowen et al. also show that the RPS is not a strong predictor of green jobs at the state level [26], while Yi shows that the RPS is strongly related to green economic development at the metro and state levels [27,28]. Yi and Feiock illustrated different political dimensions of the RPS implementation in the United States [29].

Both the RPS and MTS employ mandatory targets to improve environmental governance performance. While the former targets the energy companies, the MTS focuses on local government officials in the hierarchical bureaucratic system as its main objects. Its core issue is to deal with the possible emergence of the race to the bottom [30,31] in the local government officials’ weighing between economic growth and environment protection. To direct the officials’ attention to environmental governance, the MTS tries to adopt a series of rewards and punishment measures pertinent to the mandatory targets. Many scholars, however, still hold doubts on the effectiveness of this regulatory command-and-control method, which employs mandatory targets (rigid index) for either the local government officials or the energy companies [32,33].

In the Chinese context, the relationship between the MTS in a FYP and environmental performance has long been a highly debated academic issue. Qi indicated that the MTS set up by the 11th FYP can significantly motivate local governments and thus promote the environmental performance across China [2]. Liu demonstrated the effect of the MTS on reducing SO₂ and COD emissions [3]. Wang illustrated that the MTS explicitly regulates the administrative responsibility, and attaches rewards and punishments to the environmental performance of local governments and their officials [34]. The MTS enhances the implementation of environmental policy at the local level, because the completion of mandatory targets is incorporated in the performance assessment of local officials.

These observations are empirically supported in some studies. Hu tested the impact of the MTS on the reduction of energy intensity with time series and panel data analyses [4]. They found that the establishment of the MTS has a significant effect on the reduction of energy intensity in China. Liang tested the influence that the MTS had on the energy consumption and major pollutant emissions respectively with a panel data set at the province level in China. She indicated that the implementation of the MTS can reduce the energy consumption per unit GDP and the emissions of major pollutants in Chinese provinces, and that such effects are mainly driven by the likelihood of promotion of core leaders in the locality [5].

All the above mentioned research argues that the reason for the positive effect of the MTS on environmental performance is that the system ties the reward and punishment measures to the attainment of environmental objectives, so that it encourages the local governments and their officials to promote environmental performance improvement. Nevertheless, this attribution of causal relationship has not been examined by rigorous quantitative research. The panel data analysis conducted in the above mentioned research only demonstrates the positive correlation between the MTS and environmental performance. What is left unknown is whether such a positive relationship is driven by the reward and punishment measures of the MTS.

Some qualitative research represented by case studies has posed challenges to this attribution. It is demonstrated that the reward and punishment measures of the mandatory target can be easily evaded by the local governments during the implementation process [7,11]. The performance appraisal relevant to the mandatory target rarely influences the officials’ core interests, for example, the promotion process [8,9]. Mei provided a method for measuring environmental policy implementation by decomposing the environmental governance objectives (reduction in energy consumption per unit GDP) [6]. This research measured the correlation between the implementation of environmental policy and the policy pressures at the provincial level, and illustrated the positive effect that administrative rewards and punishments had on environmental policy implementation.

This paper addresses the limitations of the extant research and extend the literature in the following ways. First, we employ RDD estimations to detect a causal relationship between reward and punishment measures and environmental performance, for which panel data methods utilized in previous studies are insufficient. Second, we address some of the endogeneity issues shown in previous research. Third, we examine such a relationship with a longer panel data set, covering a period of three, five-year plans from 2000 to 2014. In the next section, we will provide detailed information regarding data sources and research design.

3. Data and Research Design

3.1. Variables and Measures

3.1.1. Dependent Variable: Environmental Performance

The dependent variables of this paper capture energy saving and emission reduction performance, as measured by the growth rate in three environmental performance variables. Energy consumption per unit of GDP (energy), chemical oxygen demand (COD) and the total emission of sulfur dioxide (SO₂) are selected as our three dependent variables. There are two reasons explaining our selections. Firstly, the unit GDP energy consumption, COD and SO₂ are mandatory targets regulated by the national 10th, 11th and 12th FYPs. Energy consumption per unit of GDP involves many aspects of the economy, environment and society, and is an important indicator which reflects the energy intensity embedded in the national economic development. SO₂ and COD are the major pollutants in the air and water. Therefore, the three indicators capture important aspects in environmental performance. Secondly, the three variables have been documented consistently over a long period of time. The provincial data of the energy consumption per unit of GDP date back to 2000, and the provincial data of SO₂ and COD date back to 2002. All data we use were collected from the statistical bulletin of National Bureau of Statistic of the People’s Republic of China and the Ministry of Environmental Protection. Hereafter we denote the growth rate in the three indicators as EG.

3.1.2. Mandatory Targets System

The core independent variable of this paper is whether each province establishes the MTS. The variable is measured as a 0–1 dummy variable, and herein after is marked with $D_{jt}$, where j indicates province, and t indicates year. The data are coded from the national 10th, 11th and 12th FYPs. All provincial governments set the mandatory target since 2006, therefore, $D_{jt}=1$ if $t\ge 2006$, otherwise $D_{jt}=0$.

3.1.3. Performance Accomplishment

Mandatory targets are different across provinces. In order to conduct empirical analysis, this paper calculates and analyzes target completion for each province. In order to ensure robustness of results, we construct two indicators to measure the degree of completion, which are:

$$Accomp{1}_{E,jt}=\frac{\left(E_{jt}-Ein{i}_j\right)/Ein{i}_j}{Goa{l}_{E,j}}$$

(1)

$$Accomp{2}_{E,jt}=\frac{\left(E_{jt}-Ein{i}_j\right)/Ein{i}_j}{Goal\_se{p}_{E,j}}$$

(2)

In the first indicator, the degree of completion is defined relative to the five-year energy saving and emission reduction objective $Goa{l}_{E,j}$ required in the 11th and 12th FYPs. $E_{jt}$ is the performance output of province j in year t, while $Ein{i}_j$ is the initial year (2005 for 11th Five-Year-Plan and 2010 for 12th Five Year Plan) performance of province j. When the percentage decrease of an indicator reaches or exceeds the five-year objective, it can be defined as a “completed objective” $\left(Accomp{1}_{E,jt}\ge 1\right)$. In the second indicator, we divide the five-year objective into five independent phased objectives. Assuming that each province shall complete the objective at a constant speed, when the percentage decrease of an indicator of a certain year exceeds the phased objective, it can be defined as a “completed objective” for that year $\left(Accomp{2}_{E,jt}\ge 1\right)$.

Taking the first measure for the degree of completion as an example, its time trend is shown in Figure 1. The scatter points represent each province in each year and the dotted lines show average degree of completion for all provinces in each year. In 2006–2010, the degree of completion of the 11th FYP raises gradually, wherein the average degrees of completion of energy consumption per unit of GDP and COD emission reached 100% in 2010, while that of SO₂ reached 100% in 2009. In 2011–2014, the degree of completion of the 12th FYP reached 100% earlier than in 2006–2010. The average degree of completion of energy consumption per unit of GDP exceeds 100% in 2012, and that of COD and SO₂ emissions exceed 100% in 2013.

3.1.4. Control Variables

In the regression analysis, we control for other factors that may affect provincial ability in energy saving and emission reductions. They are divided into 3 groups: governance capability, environmental pressure and local leaders’ motivation. Environmental governance capability of a local government is limited by its economic strength. Therefore, we utilize GDP growth rate as a proxy for local economic conditions. We include secondary industry share, population density and residents’ average years of education in the regressions to account for environmental pressure of each provincial region. Local leaders’ motivation for environmental governance is measured by their educational background and age. To sum up, we control for GDP growth rate, secondary industry share, residents’ average education years, population density, and the education background and age of the provincial Chinese Communist Party (CCP) secretary for each provincial region.

3.1.5. Descriptive Summary

Table 1. Summary statistics.

Variable	Obs.	Mean	Std. Dev.	Min	Max
Indicator Growth Rate (2001–2014) (unit: percentage)
Energy/GDP Growth Rate	420	−0.049	0.081	−0.471	0.715
COD Growth Rate	360	−0.004	0.096	−0.348	1.156
SO2 Growth Rate	360	0.015	0.119	−0.188	0.884
Mandatory Target System (2000–2014) (0–1 dummy)
D	450	0.600	0.490	0	1
Degree of Target Completion (2006–2014) (unit: percentage)
Ener Accomp1	270	0.960	0.908	−2.844	4.894
COD Accomp1	257 a	0.652	0.478	−0.743	1.878
SO2 Accomp1	261 b	0.664	0.959	−4.823	3.977
Control Variables (2000–2014)
Second Share (unit: percentage)	450	0.466	0.079	0.198	0.664
GDP Growth Rate (unit: percentage)	450	0.107	0.027	0.038	0.236
log Pop/Square (unit: log 1/km2)	450	8.144	0.762	6.247	9.280
Average Education (unit: Year)	450	8.188	1.027	5.300	11.836
Leader Age Initial (unit: log Age)	450	4.048	0.067	3.850	4.190
Leader: above Master (unit: 0, 1)	450	0.471	0.500	0	1

^a COD target in 11th and 12th Five Year Plan for the province of Hainan is 0%, therefore data of the degree of completion during 2006–2014 is missing. There is a similar reason for missing data of the degree of COD target completion in the province of Xinjiang during 2011–2014; ^b Missing data for SO₂ in the province of Gansu during 2006–2010 and the province of Xinjiang during 2011–2014.

represents summary statistics of key variables used in this paper. The top panel shows the growth rate of three indicators (energy consumption per unit of GDP, COD and SO₂ emissions). The middle part summarizes the mandatory target system and degree of completions. The bottom part summarizes control variables.

3.2. Research Design

The two research questions examined in this paper are: whether the MTS influences provincial environmental performance; and whether the strict reward-and-punishment measures in the MTS are the main reasons for such an impact. Therefore, the empirical research design is also divided into two phases.

The first phase is to test whether the establishment of the MTS influences provincial environmental performance in China. We put forward Hypothesis I: the establishment of the MTS has a positive effect on provincial environmental performance. Compared with existing literature, our analysis differs in several aspects. Firstly, we take three environmental performance growth rates as the dependent variables. Considering that all targets in the MTS take the form of growth rates, the measures are thus more appropriate. Secondly, we include lagged dependent variables in regressions to allow for lagged policy effects.

In the second phase, we test whether the reward and punishment measures set in the MTS influence provincial environmental policy implementation and further influence environmental performance in China. The reward and punishment measures set in the MTS return Boolean values, that is, provinces receive rewards if degree of completion reaches 100% or otherwise get punished [35]. Thus 100% completion is a critical value. This feature meets the requirement of Regression Continuity. We utilize a Regression Discontinuity Design (RDD) model which identifies the influence of the reward and punishment mechanism on environmental performance by testing whether the environmental performances of each province reduce significantly after the degree of completion reaches 100%. To sum up, we test Hypothesis II: if the rewards-and-punishment measures have impact on local governments’ environmental performances, then energy saving and emission reduction will be worse after target completeness reaches 100%. Rejection of this hypothesis implies no impact of reward-and-punishment measures on environmental performance.

4. Empirical Results

4.1. Mandatory Targets and Environmental Performance

In this section we analyze the MTS’s impact on energy saving and emission reduction performance. In the 11th and 12th FYPs, the central government assigned and allocated certain requirements for reducing energy consumption per unit of GDP, COD and SO₂ emissions to each local government. We test Hypothesis I: that the MTS has a positive effect on improving environmental performance, by estimating the following equation:

$$E{G}_{jt}=\alpha D_{jt}+\beta E{G}_{jt-1}+\Gamma X_{jt}+{\delta }_j+{\xi }_t+{\epsilon }_{jt}$$

(3)

where j and t stand for province and year, respectively. Equation (3) is estimated using Ordinary Least Squares (OLS) regression, in which we include provincial fixed effect to control for time-invariant provincial characteristics and use the year dummy to control for macro shocks. $E{G}_{jt}$ represents the growth rate of three indicators (Energy Consumption per unit of GDP, COD and SO₂) respectively. The key independent variable is a dummy variable $D_{jt}$. $D_{jt}=1$ when the mandatory target is enforced in that year, and $D_{jt}=0$ otherwise. We expect a negative value for the coefficient of $D_{jt}$($\alpha <0$), which indicates that energy consumption per unit of GDP, COD and SO₂ emissions growth rates decline after the MTS is implemented.

We include $E{G}_{jt-1}$ in the regression to control for persistence of $E{G}_{jt}$. $X_{jt}$ is a vector of other control variables, including GDP growth rate, secondary industry shares in GDP, population density, average education levels of residents, and provincial party secretaries’ age and educational background (i.e. whether they earned a Bachelor’s Degree or higher).

Before running the regressions, a series of unit root tests are performed to test the stationarity of the dependent variable. Harris and Tzavalis [36] assumes time dimension to be fixed, while another common used test, Levin-Lin-Chu [37], assumes time dimension should increase faster than cross-section dimension. In our case, time dimension is relatively small, so we utilize the Harris-Tzavalis test. Results for three dependent variables are performed with different specifications on panel mean, time trend and cross-sectional means. All specifications reject the null hypothesis of unit-root. Therefore, the dependent variable is panel stationary, and thus fixed-effect or random-effect models are applicable.

Also, Hausman tests [38] are conducted to test the specification of fixed-effect versus random-effect models. The null hypothesis that the difference of the coefficients estimated by the two specifications is not systematic is rejected, indicating the choice of a fixed-effect model (χ²(17 d.f.) = 40.87, 21.66 and 46.94 for energy, COD and SO₂ respectively). To control for heteroskedasticity, the standard errors reported are clustered at province level.

Table 2. Mandatory target system and indicator growth rate.

	(1)	(2)	(3)	(4)	(5)	(6)
Indicator Growth Rate (EG)=	Energy	Energy	COD	COD	SO2	SO2
Mandatory Target (D)	−0.0484 ***	−0.0822 ***	−0.0510 **	−0.0672 **	−0.103 ***	−0.124 ***
	(−2.65)	(−3.26)	(−2.31)	(−2.13)	(−5.46)	(−4.95)
L.EG	−0.318 ***	−0.343 ***	−0.171 ***	−0.190 ***	−0.0337	−0.0777 *
	(−6.57)	(−7.04)	(−3.12)	(−3.47)	(−0.79)	(−1.80)
Second Share		−0.0806		−0.246		−0.113
		(−0.73)		(−1.52)		(−0.95)
GDPG		−0.129		−0.0800		0.269
		(−0.54)		(−0.26)		(1.18)
log Pop		0.246 **		−0.0250		0.148
		(2.57)		(−0.18)		(1.46)
Edu		0.00473		0.00627		0.00737
		(0.31)		(0.27)		(0.44)
Age		−3.443		−14.10 *		−15.10 ***
		(−0.60)		(−1.89)		(−2.74)
Age2		0.437		1.740 *		1.866 ***
		(0.61)		(1.87)		(2.71)
above Master		0.00506		0.0180		0.0149
		(0.45)		(1.23)		(1.39)
cons	−0.00950	4.786	0.0238	28.85 *	0.0685 ***	29.36 ***
	(−0.71)	(0.41)	(1.53)	(1.91)	(4.86)	(2.62)
N	390	390	330	330	330	330
R2	0.298	0.324	0.224	0.252	0.516	0.546

t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01; Province fixed effects and year dummies included, clustered at province level.

represents the results. In Columns 1 and 2, $E{G}_{jt}$ represents energy consumption per unit of GDP. The estimate of $\alpha$ is negative and significant with or without control variables in the regression. To interpret the coefficient, after the implementation of the MTS, growth rate of energy consumption per unit of GDP was reduced by 8.22%. Similarly, in Columns 3 to 6 we find growth rate of COD emission reduces by 6.72%, and that of SO₂ emission reduces by 12.4%, since the enforcement of the MTS. Table 2 shows that provincial environmental performance, in terms of both energy saving and emission reductions, has significantly improved after the establishment of the MTS. However, it must be emphasized that the above results do not provide direct empirical evidence supporting the argument that the reward and punishment measures embodied in the MTS are the main drivers for the improvement of environmental performance.

In Columns 2, 4, and 6 we include a set of control variables that may affect provincial environment performance. The provincial secretary’s age, which represents the local leaders’ career motivation, has no significant influence on the growth rate of energy consumption per unit of GDP, but it has significant U-shape impacts on the growth rate of COD and SO₂ emissions. We find that as a provincial secretary ages, COD and SO₂ emission growth rates first decline, then increase, indicating that local leaders’ incentive to reduce pollutions first weakens then strengthens as age increases. In terms of environmental pressures, an increase in population density boosts the growth rate of energy consumption per unit of GDP. Other control variables do not show significant effects. Note that since we included provincial fixed effects in the models, all coefficient estimates capture only intra-province effects.

4.2. Sanction-and-Incentive Measures and Environmental Performance

In this section we examine whether the influence of the MTS on environmental performance is caused by the reward and punishment measures embedded in the MTS. We now test Hypothesis II: provinces’ energy saving and emission reduction speed reduces remarkably as soon as the degree of target completion reaches 100%. To test it, we first conduct a descriptive group-mean comparison t test, then employ a RDD analysis.

We separate our sample into two groups according to the degree of accomplishment for three indices respectively, then compare the performance growth rate for two groups (Group 1 for those with a degree of accomplishment less than 1 and Group 2 for others). The t test results are: compared with Group 1, Group 2 has better performance: energy growth rate is 4.55% (standard error (SE): 0.97%) lower, COD growth rate is 0.77% (SE: 0.33%) lower, and SO₂ growth rate is 1.81% (SE: 0.56%) lower. These results suggest that, after the degree of completion reaches 100%, three performance indicators become better rather than worse. It implies that rewards and punishment measures have no impact on the improvement of provincial environmental performance.

Also, we explore group comparison with graphical demonstrations. The left column in Figure 2 is a distribution graph of the degree of target completion (x axis) vs. energy consumption per unit GDP, COD and SO₂ emissions’ growth rate (y axis). In the right column for each graph, we fit a non-linear curve before and after the degree of target completion reaches 100% on the basis of the scatter points. Group 1 (degree of accomplishment less than 1) are blue triangles while Group 2 (degree of accomplishment no less than 1) are red circles. In the left column, we can see Group 2 has a lower energy growth rate compared with Group 1, and there is no systematic difference between the two groups for COD and SO₂. Further in the right column, we add a fitted curve for Groups 1 and 2 respectively, still, there exists no significant breakpoint for the three performance indicators. From the graph we observe a non-linear relationship between target accomplishment and environmental performance. When the degree of target completion reaches 100%, no significant positive breakpoint appears, leading to no support for Hypothesis II.

Next, we construct the following regression discontinuity model:

$$E{G}_{jt}=\gamma T_{jt}+\beta E{G}_{jt-1}+\theta Accom{p}_{E,jt}+\Gamma X_{jt}+{\delta }_j+{\xi }_t+{\epsilon }_{jt}$$

(4)

where j and t stand for province and year, respectively. Compared with Equation (3), we add the degree of target completion $Accom{p}_{E,jt}$ and a dummy variable $T_{jt}$ in the regression. We further include both a 1st and 2nd order of $Accom{p}_{E,jt}$ to capture the non-linear relationship between $Accom{p}_{E,jt}$ and $E{G}_{jt}$. As discussed in Section 3.1.3, we apply two different measures of $Accom{p}_{E,jt}$. $T_{jt}$ is a dummy variable, which equals 1 when $Accom{p}_{E,jt}\ge 1$, and 0 otherwise. All control variables remain the same as Equation (3) and we also include the provincial fixed effects and year dummies.

The variable we are most interested in is the coefficient ($\gamma$) for dummy variable $T_{jt}$. If $\gamma$ is significantly positive, it indicates that $E{G}_{jt}$ increases after the degree of target completion reaches 100%. That is to say, the reward and punishment mechanism provides an incentive for provinces to improve their environmental performance. On the other hand, if $\gamma$ is negative or zero, it indicates the reward and punishment mechanism provides no such incentive.

The results are shown in

Table 3. Reward and punishment measures and indicator growth rate.

	(1)	(2)	(3)	(4)	(5)	(6)
EG=	Energy	Energy	COD	COD	SO2	SO2
T	−0.0140	−0.0145	−0.00187	−0.00234	−0.00684	−0.00513
	(−1.50)	(−1.58)	(−0.42)	(−0.51)	(−0.85)	(−0.62)
E Accomp	−0.112 ***	−0.116 ***	−0.0263 ***	−0.0263 ***	−0.0104 **	−0.00896 **
	(−20.24)	(−21.15)	(−4.12)	(−3.75)	(−2.59)	(−2.14)
E Accomp2	0.0189 ***	0.0201 ***	0.00626	0.00766	0.000107	0.000344
	(11.55)	(12.29)	(1.37)	(1.55)	(0.09)	(0.27)
L.EG	−0.330 ***	−0.325 ***	−0.00760	−0.00673	0.00107	−0.00458
	(−10.14)	(−10.20)	(−0.66)	(−0.57)	(0.03)	(−0.12)
Second Share		−0.00427		0.00506		−0.0289
		(−0.04)		(0.13)		(−0.34)
GDPG		−0.414 **		0.0802		0.0980
		(−2.39)		(1.13)		(0.62)
log Pop		−0.00728		0.00885		0.0856
		(−0.08)		(0.25)		(1.05)
Edu		−0.0296 **		−0.00145		−0.00730
		(−2.58)		(−0.32)		(−0.71)
Age		−7.385 *		2.002		1.193
		(−1.90)		(1.29)		(0.35)
Age2		0.920 *		−0.252		−0.151
		(1.89)		(−1.30)		(−0.35)
above Master		0.0145 *		−0.0000933		−0.00201
		(1.93)		(−0.03)		(−0.31)
cons	−0.0234 ***	15.15 *	0.00495	−4.049	0.0178 **	−2.974
	(−3.25)	(1.90)	(1.58)	(−1.27)	(2.13)	(−0.42)
N	270	270	257	257	261	261
R2	0.801	0.817	0.488	0.498	0.401	0.409

t statistics in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01; Province fixed effects and year dummies included, clustered at province level.

and

Table 4. Robustness tests.

	(1)	(2)	(3)	(4)	(5)	(6)
EG=	Energy	Energy	COD	COD	SO2	SO2
T	0.00556	0.00299	−0.00475	−0.00593 *	−0.0157 **	−0.0150 *
	(0.57)	(0.30)	(−1.43)	(−1.71)	(−2.12)	(−1.95)
E Accomp	−0.117 ***	−0.120 ***	−0.0221 ***	−0.0214 ***	−0.00866 **	−0.00711 *
	(−20.49)	(−21.16)	(−3.16)	(−2.83)	(−2.18)	(−1.71)
E Accomp2	0.0191 ***	0.0203 ***	0.00544	0.00685	0.000151	0.000431
	(11.54)	(12.20)	(1.27)	(1.49)	(0.12)	(0.34)
L.EG	−0.324 ***	−0.320 ***	−0.0104	−0.00940	−0.0126	−0.0160
	(−9.89)	(−9.96)	(−0.90)	(−0.80)	(−0.35)	(−0.43)
Second Share		−0.00147		0.00428		−0.0102
		(−0.01)		(0.11)		(−0.12)
GDPG		−0.415 **		0.0815		0.122
		(−2.37)		(1.16)		(0.77)
log Pop		0.000539		−0.000511		0.0829
		(0.01)		(−0.01)		(1.03)
Edu		−0.0305 ***		−0.00248		−0.00755
		(−2.64)		(−0.55)		(−0.74)
Age		−6.876 *		2.408		1.317
		(−1.76)		(1.54)		(0.39)
Age2		0.857 *		−0.302		−0.166
		(1.76)		(−1.55)		(−0.39)
above Master		0.0133 *		−0.000642		−0.00162
		(1.77)		(−0.21)		(−0.25)
cons	−0.0230 ***	14.06 *	0.00626 *	−4.781	0.0235 ***	−3.213
	(−3.19)	(1.76)	(1.93)	(−1.49)	(2.70)	(−0.46)
N	270	270	257	257	261	261
R2	0.799	0.815	0.492	0.505	0.411	0.418

t statistics in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01; Province fixed effects and year dummies included, clustered at province level.

. Table 3 exhibits results with the first measure of the degree of target completion and Table 4 for results with the second measure. In both tables, estimates of $\gamma$ are statistically insignificant, reflecting that environmental performance does not backslide after targets are fully accomplished.

In Columns 2, 4 and 6 we report results with control variables. Energy consumption per unit of GDP shows a U shape as the degree of target completion increases, indicating that energy saving performance first increases then decreases with the degree of target completion. COD and SO₂ emissions growth rates decline with the degree of target completion, implying that emission reduction performance improves with the degree of target completion. An increase in GDP growth rate lowers the growth rate of energy consumption per unit of GDP. As residents’ average year of education increases, energy consumption growth rate decreases. Local leaders’ age and their educational backgrounds also influence energy consumption: energy consumption growth rate first decreases but then increases as local leaders’ age increases, and provinces with better educated leaders show a lower energy consumption growth rate. It should be noted that the goodness of fit for regression on energy is larger than that in Table 2. It is worth noting that regressions in Table 2 and Table 3 differ both at sample period (2001–2014 for Table 2 and 2006–2014 for Table 3), and at control variables (we include degree of completion in Table 3 as required by the setting of RDD).

Further, we explore the heterogeneous effect of the reward and punishment measures on environmental performance. The formation of mandatory targets follows a two-stage process. In the first stage, local governments submit proposals on energy saving and emission reduction to the central government. Then in the second stage, the central government decides to approve or reject the proposals. Upon rejection, the central government typically will assign higher targets to local governments. This process provides us with an opportunity to classify provinces into two groups. The first group includes local governments that submit relatively high targets and are approved directly by the central government, or those that do not submit any target and announce that they accept any goals set by the central government. The second group includes those local governments whose self-proposed targets are declined by the central government and have to accept a higher goal assigned by the central government. We infer that provinces in the first group tend to be more capable or willing to save energy and reduce emissions.

Following the above mentioned logic, we grouped provinces into Groups 1 and 2 according to the formation process of their mandatory targets during the 11th and 12th FYPs. Then we conducted regression of Equation (4) for each group respectively. The results are shown in

Table 5. Heterogeneity tests.

	(1)	(2)	(3)	(4)	(5)	(6)
	Type 1	Type 2	Type 1	Type 2	Type 1	Type 2
EG=	Energy	Energy	COD	COD	SO2	SO2
T	0.0209	−0.00122	−0.00812 **	−0.00445	−0.0131 *	0.00262
	(1.68)	(−0.22)	(−2.14)	(−0.18)	(−2.04)	(0.10)
E Accomp	−0.308 ***	−0.590 ***	−0.0178 **	0.0194	−0.0461	0.00329
	(−14.71)	(−4.27)	(−2.73)	(0.23)	(−1.43)	(0.02)
E Accomp2	−0.135 ***	−0.332 ***	−0.0267 **	0.00211	−0.00630	−0.109
	(−13.27)	(−6.31)	(−2.17)	(0.05)	(−1.38)	(−0.76)
L.EG	0.0234 ***	0.0418 **	0.00506	0.0183	0.0000154	−0.00465
	(11.62)	(2.45)	(0.60)	(1.15)	(0.01)	(−0.08)
Second Share	0.150 **	0.333	0.0592	−0.291	−0.0441	0.283
	(2.31)	(1.29)	(1.65)	(−0.91)	(−0.58)	(0.10)
GDPG	−0.324 *	−0.0673	0.115	0.281	−0.131	−0.370
	(−1.77)	(−0.44)	(1.36)	(1.21)	(−1.20)	(−0.20)
log Pop	−0.110	0.409 ***	−0.000798	0.0827	−0.0205	0.681
	(−0.67)	(4.46)	(−0.02)	(0.13)	(−0.31)	(0.78)
Edu	−0.0440	−0.0119	−0.000788	0.0511	−0.00944	−0.0784
	(−1.36)	(−0.81)	(−0.17)	(1.71)	(−1.65)	(−0.87)
Age	−1.844	−14.73	2.740	23.51	0.478	1.288
	(−0.57)	(−1.63)	(1.63)	(1.32)	(0.16)	(0.02)
Age2	0.235	1.811	−0.342	−2.938	−0.0594	−0.154
	(0.58)	(1.58)	(−1.61)	(−1.34)	(−0.16)	(−0.02)
above Master	0.00630	−0.0284	−0.0000857	−0.0401	−0.00135	0.0496
	(0.63)	(−1.45)	(−0.03)	(−1.46)	(−0.24)	(0.56)
cons	4.828	26.80	−5.520	−48.01	−0.647	−7.980
	(0.70)	(1.53)	(−1.57)	(−1.21)	(−0.10)	(−0.06)
N	216	54	187	43	196	28
R2	0.874	0.978	0.560	0.765	0.517	0.845

t statistics in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01; Province fixed effects and year dummies included, clustered at province level.

. The odd columns are results of Group 1 and the even columns are those of Group 2. For the first group of provinces, COD and SO₂ emissions growth rates show a significant negative breakpoint when the degree of target completion reaches 100%, while energy consumption per unit of GDP shows no significant breakpoint. For the second group of provinces, the growth rate of three performances all show no significant breakpoint when the degree of target completion reaches 100%. The above heterogeneity analysis further confirms our finding that the reward and punishment measures do not contribute to an improvement of provincial environmental performance.

5. Discussion

The empirical result of Section 4.1 provides support for the first hypothesis of this paper: “the establishment of the MTS has a positive effect on provincial environmental performance”. In the meantime, this supported the positive perception of the effectiveness of binding targets in the previous macro research literature [2,3,4]. Compared with extant studies, this paper makes three contributions. First, using the growth rate of the environmental performance indicators as the dependent variables allows us to conduct empirical analyses by matching outcome measures with the policy settings. Second, the data in this research covered the variables from the 10th Five Year Plan to the 12th Five Year Plan (2000–2013). Third, the research employs regression models based upon the results of strict unit root tests on the variables. The empirical result of Section 4.2 showed that the second hypothesis, “if the rewards-and-punishment measures have an impact on local governments’ environmental performances, then the energy-saving and emission-reduction condition will be worse after target completeness reaches 100%”, is not supported. That is, the rewards and punishment measures have no causal impact on the improvement of provincial environmental performance. This paper holds the position that the reason for this phenomenon is possibly due to two aspects: one reason being that the rewards and punishment measures with strong deterrence impact were adapted and weakened in the implementation process [7,11,39]. For example, it is found in our field interview that, although the “one-vote veto” system had been incorporated into the regulation of mandatory targets, it has changed into a veto of the province’s opportunity for honorary annual mentions, instead of a direct veto of officials’ promotion opportunities along the administrative hierarchy [40]. The reward and punishment based on the environmental performance does not have direct impact on the core benefits of officials, i.e., they cannot influence the promotion of officials in an effective way [8,9]. That is, the corresponding financial incentive pales in comparison to the investment required in the local projects.

The results of this research bridge the findings of macro level and micro level research on Chinese environmental policy. At the macro level, quantitative research of the environmental policy emphasized the positive effect of the MTS on environmental performance, while the micro level case studies pointed out that the rewards and punishment measures of the mandatory target system may not be able to motivate the local government officials. Findings in this paper demonstrated that although the environmental performance was positively related to the MTS, the reward and punishment measures of the MTS are not driving such an impact.

Then how does the MTS affect the provincial environmental performance? This paper holds the position that the positive effect may have come from the “signaling effect” of the central government policies [10]. Local governments and their officials were following the invisible power exerted by the central government, rather than being motivated by substantial rewards or punishment. As a consequence, the phenomenon of “the rich getting richer and the poor getting poorer” might occur in the environmental performance. For local governments that follow the guidance of the central government to strengthen environmental governance, they would be willing to maintain or even intensify environmental governance although they have completed the top-down goals. However, for those that are not so sensitive to the guidance of the central government, motivations to complete the required work would still be absent even though the allocated environmental governance objectives were not completed. In fact, rather than entirely relying on the top-down rewards and punishment system suggested by the prevailing perception, China’s environmental governance is a combination of central authoritarianism and local liberalism [41]. Environment policies from the central government are mainly driven by soft power such as networks [1].

The empirical results may also offer insights for the environmental policy implementation in the U.S. and EU [42], which have distinctively different environmental governance systems from China. Simple dependence on mandatory targets such as the RPS does not guarantee the ideal results suggested by the targets [24,25]. While in some states, the RPS included punishment features in its design, the results of this paper show that we need to further examine whether such punishment requirements play a role in achieving policy effectiveness. Doubts remain over whether the central government coercion works as expected in some developing countries [43].

6. Policy Implications

Several policy implications can be drawn from this study. First, in the field of environmental governance, improving the reward and punishment measures of the MTS, in ways such as the increasing the weight of the environmental indicators in the assessment of government officials, simply has little influence on the behavior of the local officials. To advance the environmental governance at the provincial level, the implementation of the reward and punishment measures of the MTS should be intensified so that the officials’ core benefits are affected by such measures.

Second, for central government, and leaders in governments at all levels with authorities over environmental protection, their political resolution should be strengthened to provide a stronger signaling effect, so that environmental governance performance may be enhanced through the informal motivation of Chinese bureaucracy.

Third, while this study focuses on the effect of centrally designed environmental accountability mechanisms, possible advances in policy design and implementation could come from provincial and local level policy innovations. As shown in the previous literature [44,45], provincial governments implement localized energy policies to stimulate the growth of green economy and promote the decarburization of their industrial infrastructures. If the central reward and punishment measures are not working as designed, perhaps provincial governments can look for more policies and management measures that are more tailored to local demands.

7. Conclusions

In the context of global environmental change, it is a critical question for environmental governance to examine whether the administrative award and punishment measures are effective in promoting environmental governance performance. Choosing the implementation of the MTS as the subject, this paper employs a fixed-effect panel data model and RDD to test whether the MTS has improved the environmental governance performance of local governments in China. The results of this research demonstrate that the MTS has a positive effect on environmental performance, however the RDD illustrates that the reward and punishment measures in the MTS have no significant effects on the provincial environmental performance. The results of this research provide a reasonable explanation to the existing gaps among the studies on the effectiveness of the MTS. This study has profound policy implications for the design and implementation of the environmental governance system in China.

Future studies are needed to further explore the dynamics of government officials’ incentives in driving environmental performance. For example, studies could examine how the political tournament has affected the implementation of environmental policies across regions and localities. Another direction of future research is to examine the adoption of localized policy designs for environmental accountability.