3.1. Methodology
This study inspects the impact of resource endowment on provincial green development. It selects industrial structure, human capital level, foreign direct investment, population density, energy consumption structure, and infrastructure construction as control variables. With the ideas provided by this research, the static panel data are produced on the basis of the above elaboration:
where
GREit denotes the level of green development of province
i in year t. An interaction term is introduced into the benchmark model. The impact of environmental regulation and provincial green development is estimated by multiplying them through decentralized processing to verify the moderating effect of environmental regulation on resource endowment and provincial green development. The regulatory effect model is constructed as follows:
According to existing studies and the above analysis, resource endowment may affect
GRE through energy efficiency. Therefore, to test how resource endowment influences provincial green development, this study examines whether energy efficiency serves a critical mediating role in the impact of resource endowment on provincial green development. In this study, the energy-efficiency factor is a mediator variable for further tests and the functional relationship of intermediary variables is determined by successive regression coefficients and the model is set as follows:
where
EFit is the intermediary variable that indicates the energy efficiency of province
i in year t. Resource endowment has an overall effect on provincial green development if coefficient
β1 is significantly positive. Resource endowment importantly affects energy efficiency if the coefficient
γ1 is significantly positive. The coefficient
γ2 represents the direct effect of energy efficiency on the high-quality development of the provincial economy after controlling for the mediating variable. Energy efficiency has a complete intermediary effect if coefficient
γ2 is insignificant. The difference and similarity between the estimated values of
γ1 ×
γ3 and
γ2 must be observed if coefficient
γ2 is significant.
Provincial green development in different provinces is not interdependent, and green development in one province may be affected by provincial green development in other provinces. Therefore, it may lead to the wrong setting of the model to ignore the spatial correlation associated with provincial green development. Based on this, spatial econometric analysis technology, which considers the spatial correlation of economic activities, is used to examine the connection between resource endowment and provincial green development. To obtain the spatial econometric model with the best fitting effect and to study if varied manners of spatial econometric models can be converted into each other, OLS-[SAR and SEM]-SAC-SDM is followed to set and test the model. The weight matrix
Wij is usually determined according to the adjacency of the space elements. The corresponding element in the weight matrix is 1 if the two regions are adjacent. Otherwise, it is 0. However, some scholars believe that the geographical-proximity matrix is not sufficient to fully reflect the objective association between provinces [
28].
In a spatial econometric model with a spatial lag term, the regression coefficient is not sufficient enough to represent the relationship between the independent and dependent variables. LeSage and Pace [
29] divided the influences of independent variables on dependent variables in the spatial econometric model into direct effects, indirect effects, and total effects, according to the varied perspectives and extent of spatial effects. Later, LeSage and Pace [
29] found that partial differential equations can compensate for the defects of point estimation in explaining spatial effects. It effectively explains the impact of random shocks on various variables to measure the direct, spatial spillover, and the total effects correctly of independent variables on dependent variables in the spatial econometric model. The calculation formula is as follows:
Convert the general form of the SDM schema to:
The above formula can be transformed into:
Upon transforming the above equation into matrix form, we can obtain:
where
m = 1, 2, 3, …, and
k represent the explanatory variable. The first matrix to the right of the equal sign refers to the partial differential matrix demonstrated by LeSage and Pace [
29]. The components on the diagonal display the average impact of the change in the
Xik variable in a specific spatial unit on the dependent variable of the unit, namely, the spatial spillover effect, direct effect, indirect effect, and total effect, which can be recorded as follows:
3.2. Variable Measurement
In this paper, five categories of variables are set: explanatory variables, explained variables, moderating variables, mediating variables, and control variables, and the meaning, metrics, and description of each variable are as follows:
- (1)
Measurement of explanatory variables
The explanatory variable is resource endowment. Presently, there is no unified method for accurately measuring resource endowment at home and abroad. In terms of the viewpoints of Li and Xu [
27], this study adopts mining practitioners to measure resource endowment. This index considers the coal, oil, natural gas, metal, nonmineral mining, and selection industries directly connected to raw materials matters and can thoroughly evaluate the economic dependence on natural resources, avoiding “measuring” the areas with highly developed regions as relatively poor resources. Provinces with more than 50% resource abundance are defined as “energy-rich areas,” while those with less than 50% are defined as “energy-poor areas.”
- (2)
Measurement of explained variables
The primary explained variable is provincial green development. Green development significantly ensures the sustainable development of the regional economy. This requires strengthening environmental governance, saving, and efficiently utilizing resources in a well-rounded way, effectively resolving resource and environmental constraints, and promoting low-carbon development. Combined with existing research results and the connotations of green development based on Wei and Li [
30], six indicators are selected to evaluate provincial green development: forest coverage, nature reserve coverage, built-up area greening coverage, sewage effluent, disposal of unusable gas, and disposal of useless solid waste. The former three indicators are used to measure provincial green environmental protection, and the latter three indicators indicate how the growth of the economy and society affects the environment. The entropy weight TOPSIS method is used to measure provincial green development in this study. The key to this approach is to use the entropy weight method to assign a weight value to each measure index depending on standardization and then to use the TOPSIS method to rank the provincial green development of each province quantitatively. The index weight value of the entropy weight method depends on the volume of information illustrated by the variation degree of each measure of index data, which minimizes the interference from human beings in the index weight assignment. The TOPSIS method quantifies the ranking by comparing the relative distance between each measurement object and the best and worst schemes, which benefits from the simple calculation and reasonable results. The entropy TOPSIS method merges the benefits of entropy weight and the TOPSIS method, making provincial green-development measurement more objective and reasonable. The step-by-step guidelines are the following:
① To eliminate the inconsistency of different measure indexes in magnitude and dimension, the range method is used to standardize each measure index
Xij in the provincial green development measurement system:
where
i represents the province and
j represents the measure index.
Xij and
Yij represent the original and standardized the measurement of provincial green-development values, respectively, and Max(
Xij) and min(
Xij) represents the maximum and minimum values of
Xij, respectively.
② This study calculates the information entropy
Xij of each measurement index
Yij in the measurement of the provincial green development system as follows:
③ This study calculates the weight
Wj of each measurement index
Yij in the measurement of provincial green development.
④ This study constructs the weighted matrix
R of the measurement of provincial green-development indicators:
where,
rij =
Wij *
Yij.
⑤ This study determines the best and worst schemes according to the weighting matrix
R:
⑥ This study calculates the sum of the Euclidean distances between each measure scheme, the best scheme, and the worst scheme.
⑦ This study calculates the relative proximity
Ci of each measurement scheme to the ideal scheme.
The relative proximity of Ci is between 0 and 1, and the larger the Ci value, the better the provincial green development. In contrast, provincial green development is worse.
- (3)
Measurement of moderating variable
Environmental regulation serves as a moderating variable. Intellectuals at home and abroad evaluate environmental regulation from the perspective of environmental regulation policies or regulatory agencies’ inspection of enterprises’ emissions [
31] and changes in pollution emissions [
32]. Relevant data on the potency of environmental regulation are difficult to obtain, so it is difficult to measure this variable. Based on the treatment method of Song and Wang [
33], this study uses the amount of contribution in environmental contamination treatment to measure the potency of environmental regulation.
- (4)
Measurement of mediating variables
Energy efficiency serves as a mediating variable. The GDP of each province is taken as the desired output variable. In this study, GDP is used as an effective output to measure the economic growth level of every province and city. The GDP deflator is converted to the actual level of GDP in terms of the year 2000 to minimize the impact of price fluctuations. The undesired output variables are carbon dioxide, sulfur dioxide, nitrogen oxide, and smoke (powder) dust emissions. The input variables are energy consumption, employment number, and capital stock. Among them, the employment number takes the mean of the beginning and end of the year. For capital stock, the year 2000 is taken as the base period and accounts for the capital stock of each province. The “perpetual inventory method” measures the fixed capital asset of provinces and cities in China.
- (5)
Measurement of control variables
Population density. The population size determines the entire region’s consumption and pollutant emission. In accordance with the viewpoint of Herzog et al. [
34], this study embraces population density as the measurement index of population size.
Industrial structure. The deviation in industrial structure among provinces and the difference in the level of total-factor energy efficiency (TFEE) of industries will affect the provincial performance of TFEE. Compared with tertiary industries with low energy consumption and high output, the development of secondary industries with high energy consumption and low output will consume more energy and the TFEE of secondary industries will be lower than that of tertiary industries. Therefore, this study selects the industrial structure index by calculating the rate of value-added secondary industry to GDP.
Energy consumption structure. The difference in the energy consumption structure of each province will inevitably lead to a difference in TFEE, and further affect it. Therefore, this study selects the percentage of coal in energy consumption as the index of energy consumption structure.
Foreign direct investment. Conversely, foreign direct investment can improve total-factor energy efficiency by directly introducing advanced foreign technology. In contrast, foreign direct investment can also change total-factor energy efficiency through technology spillover. This study uses foreign direct investment (FDI) as a measure.
Human capital. Natural resources provide a sustainable source of wealth, reducing people’s demand for the transfer of existing capital to the future; thus, resource development activities may lead to a decline in education investment returns and the requirement of educational quality. Therefore, this study adopts Li et al. ’s [
28] viewpoint to measure the level of human capital by years of education per capita.
Infrastructure construction. Improving public infrastructure can reduce transportation and transaction costs between regions. This is advantageous for foreign exchange, information communication, and the streamlining of production factors, driving the spread of technology knowledge, improving the efficiency of resource allocation and efficient use of elements, and thus contributing to the improvement of overall social and economic efficiency. Based on Bai and Bian [
35], this study uses the length of long-distance optical cables per square kilometer to characterize the provincial infrastructure environment. The index system is presented in
Table 1.
In this study, the descriptive statistics of resource endowment, provincial green development, environmental regulation, energy efficiency, and related control variables are calculated using Stata15.0, involving average, standard deviation, and minimum and maximum values, as shown in
Table 2.
Before regressing the econometric model, statistically investigating the correlation between explanatory variables is necessary. Therefore, the Pearson correlation coefficient is adopted in this study to examine the correlation and significance of all variables. The particular test results are presented in
Table 3.