Testing the Effectiveness of Government Investments in Environmental Governance: Evidence from China

Abstract

The Chinese government has taken many measures to improve the environment, such as directly investing in pollution control infrastructure, but the effectiveness remains to be tested. This paper employs the Toda–Yamamoto test and the Fourier Toda–Yamamoto test to examine the link between environmental governance investment and environmental quality in China from 2003 to 2020. PM2.5, which measures air quality, is used here as an indicator of the environment. The results reveal that environmental governance investment has notably lessened the average concentration of PM2.5 in certain regions, indicating a positive effect on environmental quality, albeit with regional variations. Taking structural breaks into consideration, the relationship between environmental governance investment and environmental amelioration is significant in a smaller number of regions. Additionally, the cross-sectional correlation is further scrutinized to assess the essential robustness of the causality between the two variables. The findings lend support to the aforementioned conclusion. These conclusions provide valuable guidance for China’s policymakers in optimizing environmental governance investments to curb pollution and achieve sustainable development.

1. Introduction

China has faced significant environmental challenges in recent decades due to rapid industrialization and economic growth. The Chinese government has introduced various environmental governance measures, such as pollutant discharge fees, environmental taxes, regional environmental supervision system, the river chiefs system, a carbon emission trading pilot, “dual control area” policy reform, and a low-carbon city construction pilot. The government also continues to invest in the construction of environmental governance infrastructure, increasing from a total investment of 106.07 billion yuan in 2000 to 24 trillion yuan in 2022. The investment in environmental pollution control is the sum of the investment in urban environmental infrastructure, the investment in industrial pollution control and the investment in environmental protection of “three simultaneous” projects. “Three simultaneous” means safety facilities shall be designed simultaneously, constructed simultaneously, and put into operation simultaneously with the main building.

China’s environment has improved significantly, and many studies have shown that the continuous governance measures are a key factor. Has the government’s investment in environmental pollution control achieved the desired effect? There is no literature to answer that question. If environmental governance infrastructure only absorbs and processes pollutants without reducing their emissions, then its effect will only be passive and short-term, and will not be a long-term determining factor for environmental improvement. Identifying this causal relationship will help the government decide whether to continue investing in environmental governance. As there are significant differences in local development in China, the effect of environmental governance investment may vary. Identifying this effect and making differential investment decisions based on it will help improve the efficiency of environmental governance investment. Furthermore, time dates may have structural breaks, and these should still be tested after dealing with structural changes. These characteristics have posed challenges for the causality test methods commonly used in the literature. The extended Fourier Toda–Yamamoto causality test can not only accurately distinguish the sample differences of causality but also determine the influence of structural changes as well. It is suitable for processing the data considered in this study. Therefore, this paper intends to use China’s provincial-level data on environmental governance investment and environmental quality to test the effectiveness of government environmental governance investment by employing the panel Fourier Toda–Yamamoto (PFTY) causality test.

The paper is structured as follows: Section 2 reviews the literature, Section 3 details the data, Section 4 presents the methodology and result analysis, and Section 5 concludes with policy implications and recommendations for further research.

2. Literature Review

Existing literature has studied the implementation effects, mechanisms, and influencing factors of various environmental regulatory measures implemented in China since the reform and opening-up.

The practice of implementing pollution discharge fees or pollution tax systems has achieved remarkable pollution control effects in developed countries. In 1978, for better sewage treatment, China also implemented policies such as pollution charges and issued pollution permits; however, the governance effect was not obvious. Moreover, pollutant emissions in some areas increased, instead of decreasing [1], because of endogenous law enforcement problems caused by the differences in the actual collection of pollution charges between provinces (the actual rate of the pollution tax designed uniformly is affected by factors such as economic development and environmental quality) and due to the inadequate enforcement caused by lower collection standards and local protectionism. In 2007, the Notice of The State Council Approving and Forwarding the Implementation Plan and Methods for Statistical Monitoring and Assessment of Energy Conservation and Emission Reduction was issued in China; it clearly stipulated that emission reduction targets should be considered as an important basis for the assessment of local governments, and the accountability and “one-vote veto” should be implemented to make pollution control a mandatory constraint for local officials. Thereafter, the results of pollution control measures have started emerging [1]. In 2007, a quasi-natural experiment based on the adjustment of pollution charge standards found that increasing pollution charge collection standards can significantly decrease the emission of pollutants per unit of industrial output and SO2 concentration in the air, resulting in an obvious emission reduction effect [2]. The significant differences in the control effects before and after the implementation of environmental policies have demonstrated that China’s environmental regulation has a mandatory restraining effect [3,4].

Environmental taxes are important environmental regulation policies in developed countries. The mechanisms and influencing factors for an environmental tax to reduce the negative externalities of environmental pollution have been analyzed based on the Pigouvian Taxes theory. Wu Jiang et al. measured the scale of environmental-related tax revenue from 2007 to 2009 in China and opined that at a certain scale, China’s environmental tax can provide certain incentives and financial support for environmental protection and reducing pollution [5]. In 2018, after the pollution charge system was implemented, China implemented the Environmental Protection Tax Law to collect environmental protection tax. Fan Qinquan et al. found that environmental tax policies can promote economic growth as well as reduce pollution levels by reducing the excessive use of energy [6]. Liu Jinke and Xiao Yiyang found that environmental taxes can improve environmental quality by boosting enterprises to implement innovative green activities to improve the utilization efficiency of fossil energy and reduce the emissions of pollutants [7]. Chen Sumei and He Lingyun conducted a theoretical deduction on the mechanism of energy tax collection to enhance economic growth and emission deduction. However, in actual investigations, the energy tax was not found to be allocated in the optimal manner to meet these two goals [8].

An environmental protection system is a basic means for the government to implement environmental regulation. However, the regulation effects of environmental laws depend on the perfection of laws as well as on how strictly they are enforced. Foreign scholars typically use the pollution tax rate and pollution governance cost as alternative indicators to measure the environmental regulation strength of a government. In addition to using environmental personnel size and environmental investment as alternative indicators to measure the environmental regulation strength of the government, domestic scholars have also conducted quasi-natural experiments on local legislations and explored the effects of environmental regulations by virtue of the difference-in-difference method [9]. Most scholars agree that stronger environmental regulations imply better environmental improvement. Li Shu and Chen Gang used the revision of the Law of the People’s Republic of China on the Prevention and Control of Atmospheric Pollution in 2000 to conduct natural experiments; they found that the revision of this law considerably decreased industrial exhaust emissions [10]. These environmental laws can exert effective stimulation and restraint on the production behavior of enterprises, which is in contrast to the findings of contemporary literature that the implementation of laws in China has not been effective [11,12].

Emission trading is one of the pollution control methods that are commonly used in developed countries. However, there are no consistent conclusions on the implementation effects of emission trading. Through experimental simulations, Schleich and Betz found that the emission permit trading system implemented in Europe positively impacts the emission reductions of small- and medium-sized enterprises. The emission reduction effect depends on the real emissions reported by enterprises [13]. Through an empirical study, Anderson et al. found that the carbon emission trading of the European Union effectively decreased the CO₂ emissions of manufacturing companies [14]. Hoffmann analyzed the impact of the European Union’s emission trading on the investment decisions of the German power industry, and they found that emission trading significantly impacts the short-term small emission reduction investments of enterprises; however, it does not impact their long-term huge emission reduction investments [15]. Based on the panel data of manufacturing in Italy, Borghesi et al. found that the implementation of the European emission trading system (EU-ETS) produced limited emission reduction results due to the loose quota issuance. Empirical evidence from China suggests that emission trading effectively decreased pollutant emissions [16]. Ren Shenggang et al. explored the emission reduction effects of China’s SO2 emission trading pilot policy since 2007; they found that SO₂ emissions decreased, and that economic growth in pilot areas was significantly higher than that in non-pilot areas [17]. The emission trading system has achieved a “win–win” for the economy and the environment. Hu et al. investigated the energy saving and emission reduction effects of the carbon emission trading pilot policy since 2011; they found that the carbon emission trading pilot policy decreased the energy consumption, as well as the CO₂ emissions, of the regulated industries in pilot areas by 22.8% and 15.5%, respectively [18].

Scholars have reported different findings when studying the regulation effects of specific measures; for example, it was reported that the policy reform in the “SO₂ Pollution Control Zone” has effectively improved pollutant reduction in some control areas [19]. The river chiefs system significantly increased dissolved oxygen in the water and achieved initial water pollution governance effects. However, the levels of pollutants in deep water were not considerably decreased [20]. The pilot policy for the construction of a “low-carbon city” considerably decreased urban air pollution by boosting enterprises to reduce emissions and upgraded the industrial structure [21]. The regional environmental supervision system has not significantly impacted the improvement of environmental quality in the study areas [22]. The environmental supervision system implemented by the central government has continuously strengthened the supervision and inspection of local governments and relevant institutions, and it has enhanced the authority of ecological environment supervision, thus significantly decreasing air pollution [23,24].

In summary, existing literature has comprehensively analyzed the implementation effects of mainstream environmental control measures globally and provided a sound basis for follow-up research. In socialist systems with Chinese characteristics, the Chinese government has not only laid the background for environmental protection policies and measures but it is also a direct executor of environmental protection. The Chinese government has been actively investing in environmental pollution governance (from 72.18 billion RMB in 1978 to 1063.89 billion RMB in 2020). The scope of investment ranges from urban environmental infrastructure to industrial pollution source control and the construction of project-related supporting pollution prevention and control facilities. However, few studies have discussed the direct environmental governance investments by the Chinese government.

Due to the great regional differences in China and the experience of many exogenous impacts and endogenous changes, there may be spatial and temporal differences in the governance effect of environmental governance investment. These spatial and temporal differences of causality have posed challenges for the causality test methods commonly used in the literature. The commonly used panel causality test is based on the overall causality test; thus, the overall causality is not significant as long as the causality of a sample is not significant. Often, some samples have causal relationships while some do not. Furthermore, time dates may have structural breaks, and these should still be tested after dealing with structural changes. The Fourier extended Toda–Yamamoto causality test can not only accurately distinguish the sample differences of causality but also determine the influence of structural changes as well. Therefore, it is suitable for processing the data considered in this study. We explore the impact of environmental governance investment on air quality by adopting the Toda–Yamamoto causality test based on Fourier function extension and used the environmental governance investment and air quality data at the provincial level in China during 2003–2020.

3. Data

This study uses the data of 30 provinces and cities in China during 2003–2020 (due to its vast area, Tibet has the minimum PM_2.5 value and environmental pollution governance investment, with a large deviation from the overall mean, and thus it was excluded from the sample. Hong Kong, Taiwan, and Macau were also excluded from the sample due to a lack of data on investment in environmental pollution governance). The data of environmental governance investment (EGI) were obtained from the China Environmental Statistical Yearbook. The most commonly used environmental quality indicators are air quality and water quality. Compared with water quality, air quality has more independent measurement data sources. Therefore, we use air quality indicators for analysis. The air quality indicators of each province take the average PM2.5 concentration data after the satellite-monitored climate data provided by Aaron et al. from Washington University in St. Louis is processed by raster processing and matched to the vector maps of 286 prefecture-level cities [25]. Complete data on government investments in environmental pollution control are available from 2003, and the latest air monitoring data are available until 2022. However, to exclude the impact of many exogenous shocks in 2021 and 2022, such as the COVID-19 pandemic, our analysis is limited to the period of 2003 to 2020. This allows us to focus on the longer-term trends and impacts of government investments in environmental governance, without the potential distortions from more recent short-term disruptions. The economic development level of each region was measured by gross domestic product per capita (gdp), and the data from the Qianzhan database were considered. Logarithms were taken for all data. Descriptive statistical analyses are presented in

Table 1. Statistics information about the variables.

Variable	gdp	EGI	PM
Obs	540	540	540
mean	10.368	9.401	3.639
min	8.216	5.886	2.293
max	12.013	11.861	4.453
p50	10.484	9.544	3.681
Std.Dev.	0.745	1.055	0.386
variance	0.555	1.114	0.149
cv	0.072	0.112	0.106
kurtosis	2.529	2.99	3.828
skewness	−0.351	−0.515	−0.706
Jarque–Bera	16.03	23.85	60.26
probability	0.00	0.00	0.00
correlation matrix
gdp	1
EGI	0.6631	1
PM	0.0609	0.4065	1

.

As shown in Table 1, EGI had the largest variance and dispersion, indicating that there is a large gap between the environmental governance investments in different regions. PM had the least data dispersion. According to the results of Kurtosis and Skewness tests, the three variables showed a leptokurtic left-skewed distribution, rather than a normal distribution. Correlation analysis showed the strongest correlation between EGI and gdp.

During 2003–2020, both the macro-economy and the air quality changed considerably; however, accurately defining the specific forms and time nodes of these changes is challenging. Fourier approximation enables an accurate analysis of the periodic fluctuations of a time series, and thus it is a common method to determine the occurrence of structural changes in a time series. Figure 1, Figure 2 and Figure 3 show the trends and Fourier fitting curves of PM, EGI, and gdp. The three indicators had a high level of Fourier goodness of fit (the goodness of fit of the fitting curve to the index reached >0.5 by setting different fitting parameters), indicating that the structural changes occurred on index variables of all provinces and cities during the analysis period.

4. Methodology and Result Analysis

4.1. Empirical Model

Because the data are not normally distributed and contain structural changes, the traditional causality test methods are not suited for stationary testing. Therefore, this paper first establishes the Toda–Yamamoto Granger causality test equation [26,27]:

$$P{M}_t={\alpha }_{1,0}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{1,1j}P{M}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{1,2j}EG{I}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{1,3j}gd{p}_{t-j}}+{\epsilon }_{1,t}$$

(1)

$$EG{I}_t={\alpha }_{2,0}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{2,1j}EG{I}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{2,2j}P{M}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{2,3j}gd{p}_{t-j}}+{\epsilon }_{2,t}$$

(2)

$$gd{p}_t={\alpha }_{3,0}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{3,1j}gd{p}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{3,2j}P{M}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{3,3j}EG{I}_{t-j}}+{\epsilon }_{3,t}$$

(3)

Further, in order to deal with the structural change of data, we add a Fourier function to the Toda–Yamamoto Granger causality test equation, as introduced by Durusu-Ciftci et al. [28].

$$P{M}_t={\alpha }_{1,0}+{\displaystyle \sum _{k=1}^n{\alpha }_{1,1k}sin(\frac{2\pi kt}{T})}+{\displaystyle \sum _{k=1}^n{\alpha }_{1,2k}}cos(\frac{2\pi kt}{T})+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{1,1j}P{M}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{1,2j}EG{I}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{1,3j}gd{p}_{t-j}}+{\epsilon }_{1,t}$$

(4)

$$EG{I}_t={\alpha }_{2,0}+{\displaystyle \sum _{k=1}^n{\alpha }_{2,1k}sin(\frac{2\pi kt}{T})}+{\displaystyle \sum _{k=1}^n{\alpha }_{2,2k}}cos(\frac{2\pi kt}{T})+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{2,1j}EG{I}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{2,2j}P{M}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{2,3j}gd{p}_{t-j}}+{\epsilon }_{2,t}$$

(5)

$$gd{p}_t={\alpha }_{3,0}+{\displaystyle \sum _{k=1}^n{\alpha }_{3,1k}sin(\frac{2\pi kt}{T})}+{\displaystyle \sum _{k=1}^n{\alpha }_{3,2k}}cos(\frac{2\pi kt}{T})+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{3,1j}gd{p}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{3,2j}P{M}_{t-j}}+{\displaystyle \sum _{j=1}^{p+d}{\beta }_{3,3j}EG{I}_{t-j}}+{\epsilon }_{3,t}$$

(6)

In the above VAR test equation, we can determine whether there is a causal relationship by performing a non-zero Wald test on the regression coefficient of the first p-order lag term of the explanatory variable. It should be noted that the p regression coefficient is an approximation χ² distribution with a degree of freedom of p. If we want to test whether the pollution control investment (EGI) has an impact on the air quality (PM), a joint test is needed for β_1,2j =0 (j = 1,…,p).

The Wald statistic may depend on the frequency parameter k in the Fourier function. It may not follow an asymptotic chi-square distribution. To deal with this statistical distribution problem, bootstrap sampling can be used to obtain the bootstrap distribution of the Wald statistic of the coefficient introduced by Becker et al. [29].

4.2. Result Analysis

4.2.1. Unit Root Test

Based on the application of the traditional ADF unit root test, this paper introduces the single-breakpoint ADF test of ZA [30] and the Fourier ADF test of EL [31]. The test results of PM and EGI show (see

Table 2. Unit root tests for environmental quality (PM).

Region	PM			d.PM
	ADF	ZA	EL	ADF	ZA	EL
Anhui	0.648	−1.695	−3.116	−4.117 ***	−6.669 ***	−5.187 ***
Beijing	1.244	−1.598	−3.326	−2.5	−4.99 **	−2.267
Chongqing	1.575	−1.679	−4.142 *	−2.172	−5.776 ***	−4.902 **
Fujian	0.872	−1.975	−3.113	−4.086 ***	−6.231 ***	−3.940
Gansu	−1.037	−1.003	−5.105 ***	−6.107 ***	−9.003 ***	−4.533 **
Guangdong	0.956	−2.251	−4.158 *	−3.348 **	−4.418	−3.917
Guangxi	0.319	−3.056	−3.986	−3.335 **	−4.76 *	−3.600
Guizhou	1.043	−2.318	−3.166	−2.629 *	−5.963 ***	−3.990
Hainan	−0.826	−3.544	−3.921	−6.022 ***	−9.054 ***	−4.595 **
Hebei	0.676	−1.564	−4.064 *	−2.935 **	−5.49 ***	−3.931
Henan	−2.973 **	−4.266	−5.057 ***	−6.126 ***	−6.905 ***	−4.559 **
Heilongjiang	0.271	−2.2	−6.686 ***	−4.221 ***	−4.466 *	−3.561
Hubei	0.718	−2.3	−3.638	−3.56 ***	−8.579 ***	−7.553 ***
Hunan	0.548	−2.137	−3.548	−3.171 **	−5.798 ***	−4.986 ***
Jilin	0.175	−1.976	−2.485	−3.768 ***	−8.079 ***	−4.457 **
Jiangsu	0.749	−0.76	−3.820	−4.293 ***	−6.864 ***	−4.770 **
Jiangxi	−1.76	−3.637	−2.903	−4.882 ***	−6.251 ***	−4.679 **
Liaoning	−1.496	−3.457	−3.038	−4.723 ***	−5.379 ***	−4.430 *
NeiMongol	−1.985	−4.633 *	−3.203	−5.266 ***	−6.509 ***	−4.840 **
Ningxia	−0.679	−3.128	−4.114 *	−4.949 ***	−4.496 **	−4.005
Qinghai	−2.129	−3.754	−4.491 **	−6.417 ***	−7.362 ***	−3.924
Shandong	0.327	−2.673	−5.078 ***	−2.561	−5.338 **	−4.472 **
Shanxi	−0.117	−3.171	−3.700	−4.205 ***	−5.738 ***	−3.074
Shaanxi	−2.375	−3.725	−3.431	−6.067 ***	−8.627 ***	−5.661 ***
Shanghai	−0.08	−2.603	−5.251 ***	−3.109 **	−5.707 ***	−4.573 **
Sichuan	1.351	−2.358	−4.427 *	−3.092 **	−6.274 ***	−3.752
Tianjin	0.314	−2.202	−3.467	−3.251 **	−5.608 ***	−4.020
Xinjiang	−1.761	−3.697	−2.639	−6.742 ***	−8.624 ***	−5.290 ***
Yunnan	0.057	−4.55	−5.197 ***	−5.184 ***	−8.229 ***	−6.129 ***
Zhejiang	0.7	−1.551	−2.721	−3.647 ***	−5.677 ***	−4.465 **

Note: The critical values for ZA test and EL test can be found in Table 2 in Zivot and Andrews [30] and Enders and Lee [31]. ***, ** and * indicate statistical significance at 1%, 5%, and 10% levels, respectively.

,

Table 3. Unit root tests for EGI.

Region	EGI			d.EGI
	ADF	ZA	EL	ADF	ZA	EL
Anhui	−2.677 *	0.124	−4.166 *	−2.951 *	−4.757 **	−4.204 *
Beijing	−1.945	−1.495	−2.004	−3.846 ***	−5.873 ***	−5.594 ***
Chongqing	−1.628	−3.625	−2.271	−3.338 **	−5.646 ***	−3.619
Fujian	−2.089	−4.107	−1.979	−4.66 ***	−5.956 ***	−3.946
Gansu	−2.074	−2.284	−2.673	−3.119 **	−5.466 ***	−3.616
Guangdong	−2.599	−4.532	−3.621	−5.042 ***	−6.316 ***	−4.740 **
Guangxi	−1.427	−3.003	−3.591	−2.102	−1.804	−1.953
Guizhou	−1.664	−2.682	−3.177	−4.581 ***	−6.3 ***	−4.847 **
Hainan	−2.679 *	−2.413	−3.093	−4.149 ***	−4.919 **	−4.348 *
Hebei	−2.061	−2.526	−4.079 *	−4.397 ***	−5.359 ***	−4.315 *
Henan	−1.428	−2.532	−4.574 **	−2.926 *	−4.672 *	−6.257 ***
Heilongjiang	−1.253	−3.704	−3.457	−5.036 ***	−5.38 ***	−4.207 *
Hubei	−2.206	−2.201	−3.935	−4.385 ***	−6.239 ***	−5.440 ***
Hunan	−1.965	−3.886	−3.465	−4.952 ***	−6.607 ***	−5.207 ***
Jilin	−2.265	−1.895	−3.483	−2.568	−3.958	1.269
Jiangsu	−2.101	−2.526	−4.096 *	−2.939 *	−4.298	−2.791
Jiangxi	−1.722	−4.142	−5.519 ***	−2.482	−3.823	−6.261 ***
Liaoning	−2.579	−2.509	−3.787	−4.445 ***	−6.581 ***	−3.968
NeiMongol	−1.595	−2.437	−4.331 *	−3.64 **	−5.092 **	−2.035
Ningxia	−1.727	−3.908	−3.216	−3.007 **	−5.774 ***	−4.192 *
Qinghai	−2.105	0.436	−3.745	−4.439 ***	−5.777 ***	−5.857 ***
Shandong	−1.674	−2.679	−4.093 *	−3.521 **	−5.569 ***	−4.716 **
Shanxi	−2.385	1.468	−4.918 **	−3.085 **	−6.803 ***	−7.628 ***
Shaanxi	−1.593	−1.088	−1.888	−2.587	−4.592 *	−3.759
Shanghai	−2.309	−0.877	−3.977	−3.765 ***	−7.012 ***	−4.418 **
Sichuan	−0.757	−4.901 **	−3.530	−5.083 ***	−6.326 ***	−4.739 **
Tianjin	−1.777	−4.823 **	−2.688	−4.591 ***	−6.241 ***	−3.922
Xinjiang	−1.269	−2.303	−6.995 ***	−3.471 **	−4.333	−5.899 ***
Yunnan	−2.339	−2.828	−3.975	−2.652 *	−4.629 *	−5.108 ***
Zhejiang	−2.371	−4.531	−4.376 **	−6.621 ***	−6.916 ***	−6.230 ***

Note: The critical values for ZA test and EL test can be found in Table 3 in Zivot and Andrews [30] and Enders and Lee [31]. ***, ** and * indicate statistical significance at 1%, 5%, and 10% levels, respectively.

and

Table 4. Unit root tests for gdp.

Region	gdp			d.gdp
	ADF	ZA	EL	ADF	ZA	EL
Anhui	−2.034	−2.396	−1.854	−2.894 *	−5.168 **	−5.051 ***
Beijing	−2.738 *	−4.477	−1.881	−2.423	−3.35	−6.061 ***
Chongqing	−3.146 **	−1.298	−3.239	−2.534	−5.085 **	−3.675
Fujian	−2.099	−1.694	−3.844	−2.569	−4.412	−3.223
Gansu	−4.956 ***	−2.989	−0.058	−1.857	−3.551	−4.032
Guangdong	−4.412 ***	−1.091	−3.540	−1.213	−4.075	−4.159 *
Guangxi	−4.718 ***	−1.416	2.565	−1.811	−4.789 *	0.347
Guizhou	−3.874 **	0.771	0.902	−1.365	−4.169	1.325
Hainan	−2.944 **	−0.871	3.491	−1.546	−3.524	1.722
Hebei	−5.114 ***	−2.465	−1.454	−2.381	−7.082 ***	−0.959
Henan	−2.423	−2.094	0.430	−3.539 **	−5.182 **	3.472
Heilongjiang	−5.357 ***	−2.06	−0.321	−2.092	−5.344 ***	0.234
Hubei	−2.654 *	−0.917	1.899	−1.859	−4.519	−2.318
Hunan	−4.833 ***	−2.524	5.579	−1.815	−4.145	3.311
Jilin	−6.284 ***	−0.932	0.375	−1.007	−3.65	0.363
Jiangsu	−4.208 ***	−2.197	−2.567	−2.628	−5.611 ***	−0.723
Jiangxi	−3.405 **	0.991	0.379	−2.476	−5.677 ***	−3.674
Liaoning	−2.957 **	−3.226	1.822	−2.436	−4.147	1.927
NeiMongol	−6.018 ***	−1.935	1.931	−1.723	−4.834 **	0.483
Ningxia	−4.109 ***	−1.52	5.554	−1.47	−4.352	1.015
Qinghai	−2.881 **	−1.508	0.466	−2.569	−4.418	−1.106
Shandong	−5.450 ***	−0.935	1.411	−1.117	−4.702 *	−0.928
Shanxi	−7.038 ***	−0.104	1.542	−1.833	−5.175 **	2.950
Shaanxi	−0.146	−4.3	5.635	−6.172 ***	−8.258 ***	0.533
Shanghai	−4.118 ***	−3.167	−4.028	−2.598	−5.874 ***	−1.798
Sichuan	−3.594 ***	−1.708	5.548	−1.787	−4.159	3.078
Tianjin	−3.081 **	1.264	0.299	−3.125 **	−6.17 ***	2.032
Xinjiang	−3.147 **	−1.053	−0.616	−2.126	−5.141 **	1.832
Yunnan	−1.481	−3.607	−0.571	−3.787 ***	−4.517	2.059
Zhejiang	−4.221 ***	−1.225	0.287	−1.528	−3.838	0.774

Note: The critical values for ZA test and EL test can be found in Table 4 in Zivot and Andrews [30] and Enders and Lee [31]. ***, ** and * indicate statistical significance at 1%, 5%, and 10% levels, respectively.

) that most provinces and cities cannot reject the null hypothesis of the existence of a unit root, but the test of the first-order difference variable rejects the null hypothesis of the existence of a unit root. Therefore, in VAR (p + d) model, the maximum order of variable (d) is 1. At the same time, the ZA and EL unit test results considering structural changes show that many samples have structural changes. The test results show that the change in gdp of most provinces and cities is a steady process. At the same time, the ZA and EL unit test results considering structural changes also show that many regions have structural changes.

4.2.2. Causality Test for Multivariate Models

Table 5. Results for causality between EGI and PM.

Region	H0: EGI Does Not Cause PM						H0: PM Does Not Cause EGI
	TY		FTY(n)				TY		FTY(n)
	W	Pboot	W	Pboot	F_F	p-val	W	Pboot	W	Pboot	F_F	p-val
Anhui	2.641	0.137	0.618	0.455	3.82	0.076	0.206	0.65	0.087	0.771	3.518	0.088
Beijing	1.37	0.545	0.232	0.626	2.996	0.115	1.798	0.48	0.107	0.742	1.843	0.228
Chongqing	8.886 *	0.067	22.94 **	0.043	11.064	0.041	1.901	0.459	4.735	0.235	4.498	0.125
Fujian	12.583 ***	0.007	2.895	0.366	0.072	0.932	0.348	0.565	0.793	0.702	5.004	0.111
Gansu	3.222 *	0.097	15.761 *	0.067	6.721	0.078	0.94	0.346	0.785	0.703	1.647	0.329
Guangdong	2.813	0.323	0.934	0.669	0.553	0.624	2.347	0.382	1.715	0.514	2.849	0.203
Guangxi	0.123	0.734	1.569	0.252	1.157	0.368	0.09	0.773	1.052	0.335	3.68	0.081
Guizhou	0.38	0.545	0.047	0.978	0.588	0.609	1.333	0.263	36.625 **	0.017	13.3	0.032
Hainan	0.232	0.649	0.732	0.728	2.815	0.205	3.733 *	0.076	0.633	0.736	0.467	0.666
Hebei	28.678 ***	0.006	7.234	0.156	2.624	0.219	0.195	0.903	6.939	0.167	9.221	0.052
Heilongjiang	0.951	0.366	0.074	0.795	0.68	0.537	0.133	0.71	0.222	0.651	11.514	0.006
Henan	1.267	0.294	0.3	0.871	1.958	0.286	0.228	0.647	13.482 *	0.085	10.334	0.045
Hubei	2.166	0.405	52.33 ***	0.009	39.242	0.007	0.429	0.805	1.199	0.599	1.541	0.346
Hunan	2.864	0.325	2.211	0.452	5.723	0.095	3.028	0.312	2.996	0.365	0.382	0.712
Jiangsu	1.627	0.226	1.95	0.47	6.478	0.082	0.021	0.888	0.103	0.956	0.562	0.62
Jiangxi	0.134	0.71	4.643	0.265	0.38	0.713	0.003	0.952	2.613	0.389	1.278	0.397
Jilin	1.72	0.466	5.624	0.219	6.425	0.082	4.159	0.233	2.51	0.396	3.117	0.185
Liaoning	3.386	0.255	1.615	0.537	7.147	0.072	1.392	0.548	6.876	0.168	8.41	0.059
NeiMongolia	3.63 *	0.087	2.557	0.402	4.783	0.117	1.752	0.22	6.012	0.179	10.714	0.043
Ningxia	0.405	0.823	2.528	0.406	16.153	0.025	1.562	0.504	2.852	0.393	2.422	0.237
Qinghai	5.116	0.161	7.86	0.158	3.011	0.192	2.798	0.327	1.298	0.593	1.53	0.348
Shaanxi	0.249	0.887	7.47	0.148	43.655	0.006	1.99	0.439	0.32	0.865	1.612	0.335
Shandong	4.773	0.18	0.108	0.952	2.148	0.264	3.777	0.254	17.661 *	0.059	16.275	0.025
Shanghai	2.295	0.168	8.775	0.133	2.079	0.271	2.003	0.192	0.586	0.75	1.061	0.448
Shanxi	2.153	0.186	1.231	0.591	3.502	0.164	0.587	0.469	0.336	0.87	1.579	0.34
Sichuan	1.47	0.246	0.057	0.818	3.079	0.11	2.223	0.169	6.173 **	0.038	4.001	0.069
Tianjin	4.477	0.208	0.463	0.795	1.386	0.375	2.179	0.4	1.856	0.487	4.608	0.122
Xinjiang	1.111	0.338	1.818	0.216	6.558	0.025	7.317 **	0.022	0.299	0.601	13.239	0.004
Yunnan	4.599 *	0.06	1.846	0.502	0.368	0.719	0.785	0.417	3.447	0.32	7.194	0.072
Zhejiang	8.036 *	0.092	8.001	0.133	2.414	0.237	7.19 *	0.096	5.279	0.207	0.652	0.582

Note: ***, ** and * indicate statistical significance at 1%, 5%, and 10% levels, respectively.

,

Table 6. Results for causality between PM and gdp.

Region	H0: PM Does Not Cause gdp						H0: gdp Does Not Cause PM
	TY		FTY(n)				TY		FTY(n)
	W	Pboot	W	Pboot	F_F	p-val	W	Pboot	W	Pboot	F_F	p-val
Anhui	0.242	0.637	1.584	0.257	3.113	0.108	0.088	0.778	2.844	0.138	3.82	0.076
Beijing	5.534	0.155	0.296	0.586	0.342	0.721	0.118	0.943	0.351	0.579	2.996	0.115
Chongqing	0.241	0.886	0.024	0.988	5.033	0.11	0.384	0.82	13.689 *	0.078	11.064	0.041
Fujian	0.233	0.631	0.34	0.849	0.188	0.838	4.128 *	0.069	1.36	0.543	0.072	0.932
Gansu	0.458	0.539	1.099	0.615	0.877	0.501	0.154	0.705	8.752	0.138	6.721	0.078
Guangdong	0.136	0.944	0.195	0.923	4.135	0.137	0.155	0.918	0.488	0.801	0.553	0.624
Guangxi	1.362	0.257	5.196 **	0.047	4.463	0.056	1.163	0.308	1.18	0.334	1.157	0.368
Guizhou	0.048	0.838	0.797	0.692	0.291	0.767	3.995 *	0.077	8.498	0.143	0.588	0.609
Hainan	3.749 *	0.098	9.559	0.118	2.3	0.248	0.511	0.489	0.863	0.681	2.815	0.205
Hebei	2.32	0.391	1.668	0.511	1.163	0.423	1.867	0.46	0.694	0.742	2.624	0.219
Heilongjiang	0.007	0.929	0.047	0.828	2.587	0.144	0.429	0.515	0.003	0.953	0.68	0.537
Henan	0.169	0.697	3.126	0.34	3.392	0.17	0.691	0.42	1.191	0.603	1.958	0.286
Hubei	0.037	0.982	0.788	0.704	0.689	0.567	6.3	0.128	24.16 **	0.035	39.242	0.007
Hunan	15.699 **	0.021	5.617	0.208	0.041	0.961	2.042	0.414	4.808	0.249	5.723	0.095
Jiangsu	0.111	0.738	1.311	0.594	0.42	0.691	0.092	0.762	2.231	0.434	6.478	0.082
Jiangxi	0.043	0.839	1.4	0.562	5.858	0.092	2.669	0.122	0.94	0.662	0.38	0.713
Jilin	17.544 **	0.025	6.69	0.169	0.186	0.84	0.959	0.668	3.782	0.304	6.425	0.082
Liaoning	1.004	0.659	15.897 *	0.06	7.702	0.066	0.248	0.877	7.666	0.151	7.147	0.072
NeiMongolia	3.458	0.1	3.56	0.311	2.186	0.26	0.589	0.47	7.15	0.154	4.783	0.117
Ningxia	152.571 ***	0	268.479 ***	0.002	5.826	0.093	2.966	0.317	3.591	0.322	16.153	0.025
Qinghai	1.312	0.554	11.015 *	0.091	9.514	0.05	1.088	0.625	3.397	0.329	3.011	0.192
Shaanxi	8.14 *	0.092	4.721	0.235	0.008	0.992	0.319	0.857	10.55	0.103	43.655	0.006
Shandong	2.742	0.343	8.945	0.127	3.065	0.188	0.798	0.683	1.662	0.499	2.148	0.264
Shanghai	0.136	0.682	1.115	0.611	0.468	0.665	0.015	0.915	2.65	0.374	2.079	0.271
Shanxi	0.347	0.567	2.669	0.409	3.779	0.151	0.547	0.492	0.54	0.769	3.502	0.164
Sichuan	0.575	0.473	0.002	0.955	1.569	0.274	0.632	0.452	0.002	0.97	3.079	0.11
Tianjin	1.99	0.448	3.626	0.314	0.975	0.472	0.285	0.87	0.977	0.67	1.386	0.375
Xinjiang	3.519	0.101	0.917	0.396	1.541	0.279	0.001	0.977	1.114	0.337	6.558	0.025
Yunnan	1.486	0.25	2.277	0.431	1.522	0.35	0.804	0.378	1.071	0.62	0.368	0.719
Zhejiang	0.846	0.675	2.238	0.454	5.279	0.104	5.693	0.151	6.784	0.179	2.414	0.237

Note: ***, ** and * indicate statistical significance at 1%, 5%, and 10% levels, respectively.

and

Table 7. Results for causality between gdp and EGI.

Region	H0: gdp Does Not Cause EGI						H0: EGI Does Not Cause gdp
	TY		FTY(n)				TY		FTY(n)
	W	Pboot	W	Pboot	F_F	p-val	W	Pboot	W	Pboot	F_F	p-val
Anhui	0.675	0.432	0.032	0.869	3.518	0.088	0.963	0.353	0.255	0.635	3.113	0.108
Beijing	0.226	0.901	0	0.994	1.843	0.228	5.635	0.161	0.983	0.338	0.342	0.721
Chongqing	0.978	0.643	0.97	0.647	4.498	0.125	1.202	0.593	3.432	0.302	5.033	0.11
Fujian	1.57	0.228	0.654	0.747	5.004	0.111	0.107	0.736	1.046	0.608	0.188	0.838
Gansu	0.859	0.385	1.863	0.475	1.647	0.329	0.969	0.334	3.11	0.337	0.877	0.501
Guangdong	30.07 **	0.012	14.638 *	0.067	2.849	0.203	0.632	0.744	1.737	0.52	4.135	0.137
Guangxi	0.041	0.838	0	0.999	3.68	0.081	0.763	0.424	0.431	0.525	4.463	0.056
Guizhou	2.857	0.117	37.031 **	0.02	13.3	0.032	0	0.996	0.613	0.766	0.291	0.767
Hainan	6.052 **	0.032	1.179	0.61	0.467	0.666	0.004	0.963	1.329	0.557	2.3	0.248
Hebei	8.318 *	0.087	14.875 *	0.065	9.221	0.052	5.218	0.179	5.007	0.24	1.163	0.423
Heilongjiang	1.233	0.302	0.076	0.786	11.514	0.006	1.12	0.327	0.29	0.613	2.587	0.144
Henan	0.325	0.557	7.087	0.168	10.334	0.045	0.59	0.474	2.182	0.427	3.392	0.17
Hubei	2.102	0.407	0.439	0.799	1.541	0.346	1.95	0.45	0.091	0.959	0.689	0.567
Hunan	0.666	0.744	0.117	0.947	0.382	0.712	11.579 **	0.048	5.902	0.205	0.041	0.961
Jiangsu	0.002	0.967	1.367	0.558	0.562	0.62	0.595	0.456	3.486	0.333	0.42	0.691
Jiangxi	0.191	0.659	3.788	0.309	1.278	0.397	0.024	0.875	1.53	0.543	5.858	0.092
Jilin	6.742	0.137	3.657	0.306	3.117	0.185	7.684	0.102	2.071	0.478	0.186	0.84
Liaoning	1.644	0.497	7.519	0.129	8.41	0.059	2.98	0.314	7.933	0.14	7.702	0.066
NeiMongolia	0.003	0.963	0.565	0.758	10.714	0.043	0.034	0.864	18.508 *	0.063	2.186	0.26
Ningxia	2.957	0.311	0.824	0.685	2.422	0.237	90.694 ***	0	262.456 ***	0.002	5.826	0.093
Qinghai	0.14	0.941	0.625	0.76	1.53	0.348	82.08 ***	0.002	219.481 ***	0.001	9.514	0.05
Shaanxi	1.398	0.524	1.796	0.504	1.612	0.335	32.63 ***	0.005	5.027	0.227	0.008	0.992
Shandong	0.249	0.889	0.937	0.667	16.275	0.025	7.832 *	0.096	13.856 *	0.073	3.065	0.188
Shanghai	0.017	0.899	0.339	0.852	1.061	0.448	0.548	0.493	1.653	0.548	0.468	0.665
Shanxi	0.534	0.493	0.112	0.951	1.579	0.34	0.051	0.821	17.554 *	0.067	3.779	0.151
Sichuan	0.272	0.626	1.891	0.192	4.001	0.069	1.64	0.237	0.457	0.53	1.569	0.274
Tianjin	1.069	0.596	0.932	0.659	4.608	0.122	0.025	0.986	1.81	0.502	0.975	0.472
Xinjiang	0.387	0.546	0.945	0.354	13.239	0.004	0.082	0.787	0.303	0.594	1.541	0.279
Yunnan	2.48	0.147	0.051	0.973	7.194	0.072	0.014	0.907	2.016	0.473	1.522	0.35
Zhejiang	0.7	0.704	0.782	0.724	0.652	0.582	3.149	0.3	8.91	0.137	5.279	0.104

Note: ***, ** and * indicate statistical significance at 1%, 5%, and 10% levels, respectively.

represent the Toda–Yamamoto (TY) and Fourier Toda–Yamamoto (FTY) causality analyses. The F-test results indicate that Fourier terms are significant in nearly 1/3 of samples, implying the samples have structural changes. This change also has an impact on the significance of causality. When structural changes in variables are not addressed, nearly a quarter of provinces and cities show that pollution control investment significantly affects air quality. After dealing with structural changes, the impact of pollution control input on environmental quality in some cities is no longer significant, such as in Fujian. In some cities, the impact of pollution control input on environmental quality has changed from insignificant to significant, such as in Hubei Province. These results show that antipollution investments significantly improved air quality in some regions, but not all regions. The effectiveness of these governance environmental controls is consistent with some recent research findings [32,33,34].

A causal test of whether air quality (PM) affects environmental governance investment (EGI) showed a structural change in the relationship in the same one-third of samples. In three regions, air quality significantly affected pollution control investments. After accounting for structural changes, air quality in four regions significantly affected pollution control investments. These results show that air quality status is not the main determinant of local government pollution control investment. The government’s investment decisions in environmental governance are influenced by various factors. These factors include the following: (1) Local fiscal capacity—differences in fiscal strength in different regions may lead to different levels of government investment in environmental governance. Regions with relatively strong financial resources may have more resources invested in environmental protection. (2) Economic development needs—some areas may prioritize economic growth over environmental governance. Environmental investment is adjusted based on the local government’s development stage and economic goals. (3) Central policy orientation—the central government’s policy orientation and incentive mechanism may also impact local government environmental investment decisions. If the central government provides high policy priority to environmental governance, local governments will increase their corresponding investments. (4) Interest-related games—different interest groups, such as enterprises and residents, may affect a local government’s environmental investment decisions. The government needs to balance the interests of all parties.

Table 6 reports the causality test results between PM and gdp. The TY test shows unidirectional causal relationships from PM to gdp in Hainan, Hunan, Jilin, Ningxia, and Shaanxi. The unidirectional causal relationships from gdp to PM are significant in Fujian and Guizhou. However, when structural breaks are taken into account, causality relations are detected for some new samples. According to the FTY testing procedure, unidirectional causality from PM to gdp is also found in Guangxi, Liaoning, and Qinghai. For Chongqing and Hubei, gdp is a cause of PM. Our results are consistent with previous analyses [4,35]. Some regions have restricted the development of highly polluting industries in order to protect the quality of the environment, or have “shut down and transferred” the existing highly polluting industries in response to the final assessment of the completion of the binding energy conservation and emission reduction indicators by the central government in 2010 [4], thus affecting the development of the local economy. In addition, air pollution will also lead to a decrease in labor supply and an increase in medical expenditure, thus affecting GDP. Chen and He found that the GDP loss caused by air pollution in 2007 reached 361.468 billion RMB [32]. For the remaining areas, there is no causality in any direction between PM and gdp.

Table 7 reports the causality test results between gdp and EGI. The TY result indicates unidirectional causality from gdp to EGI for Guangdong, Hainan, and Hebei. The FTY test results confirm a unidirectional causality from gdp to EGI in Guizhou. Most local governments do not determine the investment in pollution control according to the development of economics. This is related to the transformation of China’s environmental regulation approach from the former command and control type to the market type [4,7]. In the process of transformation, the government has paid more attention to the use of market mechanisms to form constraints and incentives for enterprises to reduce emissions, which has alleviated the passive environmental control situation that the local government mainly dealt with after pollution. According to the data released by the National Bureau of Statistics, the proportion of the total environmental governance investment in China’s GDP gradually increased from 1.14% in 2001 to its highest, 1.49 percentage points, in 2008, and then gradually decreased to 0.9 percentage points in 2019 and 1 percentage point in 2020. At the same time, the proportion of the national total environmental governance investment in local fiscal revenue fluctuated from 16.55% in 2000 to 9.05% in 2019 and 10.6% in 2020.

The causality from EGI to gdp is supported by the Toda–Yamamoto test results in Hunan, Ningxia, Qinghai, Shaanxi, and Shandong. It also supported by the Fourier Toda–Yamamoto test results in NeiMongolia and Shanxi. Pollution control investment will crowd out other fixed asset investments, thus affecting economic growth. The smaller the financial scale of the region, the more significant and huge the impact.

In sum, a causal relationship among the three variables was found in nearly 1/5 of the samples, and nearly 1/3 of the samples showed that there was structural change in the causal relationship between the variables. Moreover, after adding the Fourier function to deal with the structural transfer, the causal relationship between the variables changed in all provinces and cities.

4.2.3. Causality Test for Bivariate Models

Cross-sectional dependency is an important feature in cross-region analysis in a unified market context. Because neglecting cross-sectional dependencies can lead to serious deviations and scale distortions in analysis, it is important to control cross-sectional dependencies across samples, as per Pesaran [36].

The cross-regional movement of air pollutants will also lead to regional correlation, so pollution control will also have spillover effects. There is a strong correlation between environmental pollution governance, economic growth, and air quality. The causal relationship between any two variables will be affected by the third variable. Consequently, we used many methods to test cross-sectional correlation, such as Lagrange multiplier (LM) tests [37], cross-sectional dependence (CD) tests [38], and bias-adjusted LM (LMadj) tests [39]. In

Table 8. Results from cross-sectional dependency tests.

	EGI-PM		EGI-gdp		gdp-PM
	Statistics	p-val	Statistics	p-val	Statistics	p-val
LM	4423	0.00	3174	0.00	4732	0.00
CD	63.98	0.00	51.29	0.00	67.01	0.00
LMadj	307.1	0.00	209.7	0.00	325.9	0.00

, the results show that the null hypothesis of no cross-sectional dependence across regions is strongly rejected.

Table 9. Results for causality between EGI and PM from the panel Granger causality test.

Region	H0: EGI Does Not Cause PM					H0: PM Does Not Cause EGI
	Wald	p-val	Bootstrap CVs			Wald	p-val	Bootstrap CVs
			1%	5%	10%			1%	5%	10%
Qinghai	0.543	0.908	444.985	149.181	89.191	1.904	0.821	347.077	146.433	88.027
Ningxia	13.753	0.437	371.051	119.208	75.729	32.876	0.255	328.922	117.545	72.574
Hainan	2.14	0.736	396.36	138.528	77.439	6.272	0.654	389.989	151.005	98.878
Gansu	1.237	0.816	567.505	143.15	89.641	22.708	0.296	256.809	112.373	63.337
Jilin	5.183	0.635	390.429	150.511	82.869	5.818	0.635	417.5	153.159	93.448
Heilongjiang	84.454	0.154	605.391	249.615	124.018	16.146	0.517	380.002	178.109	105.024
Tianjin	3.358	0.675	518.689	159.605	87.214	7.494	0.482	232.756	84.34	47.627
Xinjiang	4.006	0.68	334.161	108.588	68.656	65.622	0.184	574.46	209.107	127.858
Guizhou	14.532	0.397	322.695	119.176	75.322	20.149	0.318	223.797	91.057	58.36
NeiMongol	5.603	0.614	339.604	131.073	72.429	35.267	0.265	505.413	151.551	93.667
Shanxi	0.063	0.965	330.489	124.772	86.341	27.079	0.296	321.249	142.969	83.016
Guangxi	3.468	0.634	217.843	95.523	63.338	9.214	0.48	246.285	88.129	55.792
Yunnan	8.131	0.63	481.367	178.776	113.397	19.491	0.333	312.287	93.178	64.754
Liaoning	1.909	0.79	382.224	147.648	93.873	8.737	0.549	382.616	140.742	83.435
Chongqing	21.917	0.336	341.271	127.902	80.431	1.409	0.78	236.807	132.408	82.706
Jiangxi	16.726	0.388	276.585	117.338	78.933	38.566	0.376	354.803	147.33	96.673
Shaanxi	12.375	0.424	292.275	130.184	70.647	29.037	0.239	572.523	207.402	125.514
Beijing	4.417	0.705	417.859	175.442	106.09	35.027	0.221	360.761	173.167	115.372
Hebei	3.459	0.659	239.543	101.214	56.622	1.608	0.325	416.484	160.808	90.692
Anhui	10.213	0.599	699.78	204.467	118.229	15.772	0.192	365.559	147.305	100.266
Shanghai	0.527	0.905	438.504	177.873	107.281	28.166	0.712	420.478	193.744	132.403
Hunan	23.352	0.365	518.412	145.84	90.158	27.133	0.333	451.535	174.533	109.744
Fujian	22.908	0.265	383.68	130.827	59.685	48.348	0.096	264.869	139.226	88.802
Hubei	30.022	0.245	274.397	113.219	75.564	20.005	0.807	498.753	190.053	121.641
Sichuan	27.767	0.291	244.588	111.088	64.924	16.147	0.830	539.964	228.567	120.924
Henan	11.726	0.527	398.782	170.31	107.804	47.217	0.145	359.037	155.293	99.866
Zhejiang	2.951	0.716	404.711	121.567	76.727	0.130	0.079	330.384	171.946	101.234
Shandong	0.081	0.967	422.802	176.499	113.237	43.792	0.123	804.413	249.696	151.121
Jiangsu	18.441	0.454	965.362	290.535	145.785	54.160	0.728	429.26	182.443	120.209
Guangdong	31.353	0.246	334.214	116.015	71.611	4.747	0.854	292.875	136.809	84.403

Notes: 1. In the estimation of the seemingly uncorrelated regression (SUR) model, the best combination of lag terms of order 1 to 4 is selected according to the Schwarz–Bayes criterion. Simulations based on 5000 samples give critical values for the parameters 10%, 5%, and 1% significance. The following results are also obtained in this way. 2. The provinces and cities under the region are ranked according to their total GDP in 2021 from low to high.

,

Table 10. Results for causality between PM and gdp from the panel Granger causality test.

Region	H0: PM Does Not Cause gdp					H0: gdp Does Not Cause PM
	Wald	p-val	Bootstrap CVs			Wald	p-val	Bootstrap CVs
			1%	5%	10%			1%	5%	10%
Qinghai	25.169	0.389	327.591	154.837	107.584	2.393	0.63	303.691	93.393	46.587
Ningxia	220.02	0.037	442.132	181.956	106.503	11.44	0.229	155.343	46.626	29.355
Hainan	18.596	0.505	546.908	187.724	118.295	21.072	0.156	233.99	63.726	33.392
Gansu	12.344	0.5	450.405	128.654	78.194	4.679	0.554	293.466	104.047	51.956
Jilin	73.146	0.112	337.773	127.406	79.577	1.369	0.732	272.403	99.669	52.593
Heilongjiang	0.572	0.913	531.074	233.178	134.146	0.323	0.88	342.768	128.471	72.104
Tianjin	55.325	0.145	327.586	131.843	74.141	12.104	0.44	474.638	164.882	97.053
Xinjiang	52.991	0.285	656.362	233.55	138.238	0.592	0.806	236.692	75.804	43.746
Guizhou	77.778	0.1	329.97	114.133	76.505	31.31	0.154	202.71	84.37	50.447
NeiMongol	0.557	0.917	432.557	149.778	92.193	4.295	0.663	382.6	176.973	106.255
Shanxi	17.818	0.44	400.463	158.265	93.404	9.967	0.474	518.199	144.588	87.455
Guangxi	5.673	0.67	421.949	158.673	95.446	18.642	0.196	166.206	56.318	36.268
Yunnan	12.846	0.432	288.113	121.445	73.282	30.744	0.166	277.144	91.06	51.644
Liaoning	0.815	0.858	266.094	139.514	84.916	8.205	0.563	1160.43	190.581	112.666
Chongqing	25.846	0.317	449.074	162.689	94.951	31.988	0.103	203.857	57.894	32.18
Jiangxi	4.785	0.717	446.097	166.383	110.689	15.405	0.222	327.762	86.567	37.551
Shaanxi	58.09	0.164	389.79	147.042	92.456	6.771	0.354	296.183	94.527	44.064
Beijing	61.62	0.233	595.287	210.843	129.101	1.558	0.747	615.417	196.414	111.088
Hebei	2.424	0.84	589.017	243.422	143.705	2.213	0.693	708.227	171.127	84.486
Anhui	11.585	0.48	314.131	130.425	80.812	6.859	0.426	485.313	98.256	52.886
Shanghai	5.621	0.735	909.742	331.893	159.712	5.648	0.647	728.244	236.885	114.749
Hunan	1.597	0.83	436.115	198.953	124.011	29.241	0.173	371.235	104.587	57.763
Fujian	9.608	0.508	322.299	117.243	74.927	29.027	0.099	234.504	59.562	28.87
Hubei	56.579	0.244	354.025	163.491	121.738	15.88	0.191	221.999	53.748	31.624
Sichuan	15.883	0.564	509.396	232.094	130.665	11.362	0.279	237.308	93.27	43.46
Henan	2.003	0.777	334.797	132.877	80.444	6.516	0.365	198.922	79.929	41.191
Zhejiang	21.679	0.447	345.132	163.876	102.532	5.369	0.438	358.204	96.005	44.067
Shandong	102.771	0.144	767.215	246.98	141.013	2.596	0.695	541.398	207.32	99.979
Jiangsu	75.194	0.129	340.788	137.589	92.653	3.611	0.472	258.257	58.418	30.885
Guangdong	9.658	0.621	486.985	157.01	104.887	36.886	0.143	390.349	108.172	53.468

and

Table 11. Results for causality between gdp and EGI from the panel Granger causality test.

Region	H0: gdp Does Not Cause EGI					H0: EGI Does Not Cause gdp
	Wald	p-Val	Bootstrap CVs			Wald	p-Val	Bootstrap CVs
			1%	5%	10%			1%	5%	10%
Qinghai	0.207	0.883	307.055	89.138	49.148	8.489	0.676	446.903	179.302	115.293
Ningxia	6.493	0.384	261.078	72.52	41.08	371.526	0.01	367.009	137.734	81.097
Hainan	1.632	0.652	218.843	63.872	37.021	25.737	0.403	464.874	187.079	126.385
Gansu	0.827	0.734	225.685	70.578	39.121	3.448	0.749	456.262	196.259	114.201
Jilin	0.158	0.92	291.032	95.271	53.703	201.42	0.028	325.61	140.029	92.024
Heilongjiang	0.001	0.991	310.563	129.154	70.192	28.404	0.462	719.596	261.154	160.955
Tianjin	4.59	0.483	383.388	86.257	48.729	88.563	0.149	451.087	186.367	117.254
Xinjiang	5.259	0.46	285.633	73.692	43.048	12.108	0.452	318.621	120.568	81.159
Guizhou	18.793	0.172	291.574	67.527	34.27	210.834	0.043	484.388	194.997	125.773
NeiMongol	14.429	0.399	390.576	137.828	74.062	7.568	0.653	395.494	179.829	111.381
Shanxi	11.925	0.367	604.948	143.169	72.701	10.991	0.482	324.008	114.103	74.747
Guangxi	0.002	0.99	176.969	68.867	39.248	136.204	0.065	373.669	160.355	100.258
Yunnan	4.579	0.491	244.41	77.19	42.501	21.333	0.533	561.947	204.887	147.306
Liaoning	1.721	0.749	501.112	153.108	96.583	27.794	0.32	345.003	144.533	96.614
Chongqing	0.955	0.728	196.727	62.709	35.706	22.677	0.394	331.174	152.104	93.459
Jiangxi	20.375	0.178	178.271	68.193	37.527	12.964	0.529	374.535	168.125	104.482
Shaanxi	1.738	0.708	456.318	150.208	77.096	1.689	0.758	265.268	122.027	74.522
Beijing	6.979	0.070	430.493	185.48	113.454	21.649	0.456	483.409	200.887	132.037
Hebei	24.306	0.216	335.027	127.744	64.952	1.847	0.762	326.039	131.067	81.162
Anhui	41.184	0.356	204.686	78.752	42.254	0.668	0.875	498.269	210.603	118.461
Shanghai	76.822	0.296	468.727	197.791	114.852	12.518	0.595	653.237	213.14	127.558
Hunan	40.202	0.882	355.087	139.36	88.002	60.661	0.246	525.776	207.275	127.756
Fujian	63.746	0.709	245.352	101.348	60.831	162.683	0.03	361.401	110.924	70.096
Hubei	94.248	0.189	215.072	92.084	52.76	190.774	0.047	451.622	179.618	107.334
Sichuan	18.077	0.032	224.044	81.831	40.007	97.488	0.115	336.551	167.218	110.126
Henan	67.758	0.936	261.908	89.792	47.802	7.357	0.65	888.905	230.129	123.982
Zhejiang	15.103	0.602	298.256	111.62	59.78	160.88	0.055	469.971	170.041	106.133
Shandong	0.664	0.398	424.478	114.38	66.486	371.054	0.024	563.325	224.182	140.422
Jiangsu	99.842	0.570	223.083	66.16	34.609	190.752	0.086	1211.25	339.569	175.219
Guangdong	8.695	0.924	376.999	120.268	68.528	1.086	0.882	456.041	171.411	108.428

show the results of the panel Granger causality test, which controls for cross-sectional dependency, proposed by Kónya [40]. According to Table 9, there is no unidirectional causality from EGI to PM in any sample, while the causality in an opposite direction is supported in Fujian and Zhejiang. Environmental pollution has caused the government to invest in pollution control; however, environmental governance investments have not significantly improved air quality. Environmental governance measures have led to some improvement in air quality in urban centers in China, but they have been less effective in addressing air pollution in rural areas, where a significant proportion of the population still relies on solid fuels for cooking and heating [41]. Environmental governance investments may also face such a situation, so the overall effect is not significant.

In Table 10, the results indicate that there is unidirectional causality between PM and gdp in two regions. For Ningxia, we find evidence of causality from PM to gdp and from gdp to PM in Fujian. For the remaining samples, the results indicate that the null hypothesis of PM and gdp not having causality regarding each other is accepted. In other words, environmental pollution has not affected the development of the local economy, and economic development is not a significant factor affecting environmental pollution.

In Table 11, there is unidirectional causality from gdp to EGI in Beijing and Sichuan. Economically developed regions have more financial resources to invest in environmental protection. The results also indicate unidirectional causality from EGI to gdp in eight samples. EGI has a negative effect on gdp in Ningxia, Jilin, Guizhou, and Guangxi. Investment in environmental governance will crowd out other investments and affect economic development in underdeveloped areas. At the same time, EGI has a positive effect on gdp in Fujian, Zhejiang, Shandong, and Jiangsu. In developed regions, a win–win situation of air quality improvement and economic growth driven by environmental governance investment may be formed, which is consistent with the potential hypothesis [42].

5. Conclusions

In this paper, an extended Fourier Toda–Yamamoto causality test has been introduced and used to examine the causal relationships between environmental pollution governance and environmental quality in China. In particular, the method has skillfully handled regional differences and the structural shifts in time and illustrated more detail about the causality. Our findings show that environmental pollution governance has a significant influence on environmental quality, with regional differences. The structural shifts over time of variables in some provinces and municipalities have also affected the relationship between environmental pollution governance and environmental quality. Finally, a causality test method that controls for cross-sectional correlation was used for robustness testing. The results also support a causal relationship between environmental pollution governance and environmental quality. These conclusions are an important supplement to studies on the treatment of structural changes and regional differences.

According to the aforementioned conclusions, we can take measures to improve the environmental control effect of environmental governance investment. (i) In general, investment in environmental governance has a positive effect on air quality, and the government should continue to strengthen investment in environmental governance. Pollution control has a strong positive externality, which belongs to the field of market failure. Strengthening the role of government is a beneficial supplement to market failure. (ii) The regional distribution of environmental pollution control investment should be optimized and adjusted. The results show that there is significant regional heterogeneity in the effects of environmental pollution investment on air quality. Therefore, it is necessary to increase investment in areas with obvious effects and reduce investment in areas with poor effects. (iii) In order to improve the effect of environmental pollution control investment, we should adjust the investment field of environmental pollution control investment. From the previous investments in the absorption and treatment of pollution emissions, the main investments in pollution prevention and control and technology reduction can advance.

The panel causality test method adopted in this paper has well captured the regional heterogeneity of causality, and the extended Fourier Toda–Yamamoto causality test has handled the structural shifts very accurately, which provides a more advanced method for the causal testing of panel data. There are many problems that need further study. First, there is a strong spatial spillover effect of environmental governance investment. This paper simply verifies the cross-sectional correlation of the data and does not further measure the spatial correlation; second, there is a strong lag effect in environmental governance investment. If the lag term is introduced for dynamic causal test, it will be more explanatory to reality. Third, we focus only on the causal links among environmental governance investment, GDP, and PM variables. In fact, there are several variables that might affect PM in reality, such as green finance, technical innovation, regional tree planting acres and green roofs [43], and so on.