Analyzing China’s provincial pollution and its influencing factors: A 1 spatial analysis 2

: In-depth analyses of the spatial heterogeneity in pollution, and the causes of 8 differences are of great importance for contributing to provide reference for reduction 9 policies. However, a spatial analysis of the existence and mechanism of China’s 10 pollution is still ignored. Using the province-level data of thirty provinces in China over 11 2005-2017, this paper constructs a spatial Durbin model (SDM) to empirically address 12 the existence and spatial transmission mechanism of pollution. The main results are as 13 follows: first, China’s pollution shows significant characteristics of spatial dependence 14 and clustering from global and local perspectives, indicating that the existence of spatial 15 autocorrelation in pollution across regions. Second, both per capita GDP and 16 urbanization have positive impacts on pollution, but the impacts of environmental 17 regulation and FDI are insignificant. Third, urbanization not only directly influences 18 pollution, but also indirectly influences pollution. Our analysis provides valuable 19 information for developing policies to effectively alleviate pollution.


Introduction
the novel characteristics of this paper, suggesting everything is more closely related to 66 each other in spatial distribution (Tobler, 1970). Spatial econometric models consider 67 both the effects of influencing factors and spillover effects with neighboring regions. In  In summary, previous scholars have extensively focused on pollution and its 75 influencing factors. However, there are still some research gaps. Extant researches 76 ignore the existence and mechanism of pollution from a spatial perspective.

77
Undoubtedly, an accurately comprehensive understanding of the spatial transmission 78 mechanism of pollution through a spatial econometric approach is a scientific basis for 79 promulgating environmental policies to effectively control pollution. Regional 80 heterogeneity and spatial correlation are essential characteristics affecting the impacts 81 of driving factors of pollution. Due to the presence of spatial interconnection, the local 82 pollution may exert spillover effects on the pollution of adjacent regions through 83 diffusion or radiation (Pan et al., 2015). Therefore, the environmental pollution of 84 various regions are both interrelated and distinct. Whereas the spatial dependence and 85 spatial correlation of economic units may exist among adjacent regions, ignoring 86 significant spatial spillover effects would lead to bias in estimation results. On one hand, 87 the exchange of resources or technology between regions may lead to the spatial 88 spillover and diffusion effects of environmental pollution of one area, which affects 89 neighboring areas. On the other hand, the gravitational effects of spatial units can lead 90 to spatial correlations in pollution.

91
To fill these gaps, using a province-level data of thirty provinces spanning from 92 the year 2005 to 2017, this paper explores the influencing factors on China's pollution, 93 specifically to test the existence and spatial transmission mechanism from direct and 94 spillover effects perspectives. More importantly, we provide a corresponding tailored 95 strategy that can effectively examine the spatial spillover effects. This mostly differs 96 from extant literature that hardly focuses on the spatial spillover effects of pollution.

97
Therefore, considering the similarity of economic units among regions (Tobler, 1970), 98 spatial effects cannot be ignored in policy effects. By performing these analyses, we 99 expect to offer empirical evidence for the existence of spatial agglomeration in pollution,  The structure of the paper is as follows. Section 2 describes the methodologies.   Before performing spatial econometrical analysis, it is essential to explore the spatial 126 autocorrelation of core variables. We use both the global and local spatial 127 autocorrelation tests for core variables. The calculation formulas are denoted as Eqs.
represents a spatial weight matrix.  respectively. is a random error term.

139
First law of geography indicates everything is more closely interrelated to each 140 other in spatial distribution (Tobler, 1970). The results of the traditional panel models where is spatial autoregression coefficient. is the spatial lag term, denoting the 149 effect from the independent variables on the explained variables.  constructed (e.g., adjacent, geographical distance, and geography-economy weight 164 matrices).

165
The adjacent matrix is based on the geographic location between the units, which 166 is calculated as Eq. (6): The geography-economy matrix is based on both geographical distance and spatial 169 economic linkages, which is calculated as Eq. (7): Where ̅̅̅̅̅̅ refers to the average actual GDP of the region .

172
The geographical distance matrix is based on the latitude and longitude 173 coordinates of the regions, which is calculated as Eq. (8):

Decomposition effects 176
To consider the potential spatial spillover effects, the increase of explanatory 177 variables will not only bring about the increase of local pollution, but also exert its Formally, Eq. (9) can be rewritten as: As displayed in Eq. (10), the direct, total, and indirect effects can be rewritten as:   The economic structure in various regions leads to significant differences in

273
The estimation results for the SDM model with matrices W1 and W2 are shown in 274   Table 2. It is noteworthy that R 2 are relatively high, which suggests better fitting 275 models. Thus, an analysis of the SDM model will then illustrate its driving factors. it is vital for performing spatial econometric models, considering spatial effects, to 283 analyze the driving factors affecting pollution and, to examine the spatial spillover 284 effects.

285
As seen in Table 2, TEC exerts a negative impact on pollution with matrices W1 286 and W2, indicating that a higher technological level will result in less pollution. One  Table 3.
318 Notes: *, **, and *** respectively represent significance at 10%, 5%, and 1%. 320 Table 3, the first column displays the direct effects. The direct effect 321 of TEC is significantly negative with matrices W1 and W2. This indicates that the 322 technology is further improved, the industrial structure has been gradually upgraded 323 and optimized, and thus reducing the pollution. By using innovative clean technologies, 324 the cost of producing and using clean energy is greatly reduced. Therefore, a wider use 325 of clean energy may be possible, which significantly decreases pollution. The direct 326 effects of PGDP and UR are significantly positive with matrices W1 and W2, indicating 327 that the development of and economic and urbanization increase pollution. However, 328 the direct effect of FDI is not significant with matrices W1 and W2. The direct effect of 329 RE is significant with matrices W2 whereas not significant with matrices W1.

330
In column 2 of Table 3   In column 3 of Table 3 shows the total effects. The total effect of PGDP positively 340 influenced pollution with matrices W1 and W2. The total effect of RE is also positive 341 and significant with matrices W1. However, the total effect of UR negatively influenced 342 pollution with matrices W1 and W2. FDI is also negative and significant with matrices 343 W1. Notes: *, **, and *** respectively represent significance at 10%, 5%, and 1%. 346

Robustness test 347
To further test the validity of the above results, this paper utilizes the geographical 348 distance matrix and, the result is shown in Table 4. As shown in Table 4  Notes: *, **, and *** respectively represent significance at 10%, 5%, and 1%.    a cross-regional joint mechanism. The local governments should establish a cross-427 regional joint mechanism and stronger regional cooperation to combat pollution. Since