Frontiers in Business, Economics and Management 
ISSN: 2766-824X | Vol. 13, No. 1, 2024 

 

69 

Research	on	the	Influence	and	Spatial	Effect	of	Carbon	
Trading	Market	on	Green	Innovation	Efficiency	of	
Industrial	Enterprises	
Yingcheng Wang* 

College of Economics and Management, Tongji University, Shanghai, China 
*Corresponding author: wangyingcheng2021@163.com 

 

Abstract: Based on the panel data of 30 provinces and cities in China from 2005 to 2020, a two-stage value chain model of 
green innovation of Chinese industrial enterprises is constructed and the green innovation efficiency of industrial enterprises is 
measured, in which coastal cities such as Guangdong, Beijing, and Shanghai have the highest innovation efficiency, and inland 
cities such as Ningxia, Inner Mongolia, and Gansu have the lowest efficiency. On this basis, a spatial double difference model is 
applied to reveal the effect of carbon trading market policies on urban green innovation efficiency and its spatio-temporal 
heterogeneity. The results show that the carbon trading market policy significantly affects the green innovation efficiency of 
industrial enterprises, and at the same time, the level of economic development, the level of foreign investment, and the industrial 
structure of each province and city also have a positive effect on the green innovation efficiency. There is also a spatial spillover 
effect of the carbon trading market policy, which not only has a significant incentive effect on the green innovation efficiency of 
the pilot cities, but also has a positive promotion effect on the green innovation efficiency of the neighboring cities. At the same 
time, the support of provincial and municipal governments significantly affects the spillover effect of carbon market policies. 

Keywords: Carbon trading market; green innovation efficiency; spatial double difference; super-efficiency DEA. 
 

1. Introduction 
Climate change is a major challenge facing the world, and 

addressing climate change has a bearing on both China's 
domestic and international situations, as well as on the overall 
situation of development and the long term, and is not only an 
important tool for promoting high-quality economic 
development and ecological civilization, but also an 
important area for participation in global governance and 
adherence to multilateralism. The carbon emissions trading 
market, as a market-based environmental management tool, 
has been launched since June 2013 in seven pilot provinces 
and cities: Beijing, Tianjin, Shanghai, Chongqing, Hubei, 
Guangdong and Shenzhen. The government carries out 
carbon emissions trading pilot work, using market means, 
actively inducing enterprises to carry out green technological 
innovation, while the technological progress of enterprises is 
a key factor in carbon dioxide emission reduction, reducing 
future carbon emissions, can effectively realize the low-
carbon development of enterprises, and promote the green 
development of China's economy. In this context, assessing 
the actual policy effect of carbon emissions trading on 
enterprise green innovation has become a hot issue of concern 
for policy makers and academics. Under the regional 
coordinated development policy, the spatial correlation 
between different parts of the country should not be neglected, 
and it is of great significance to understand the current 
situation of China's carbon trading market and the 
development of green innovation of industrial enterprises as 
well as the spatial and temporal evolution patterns, and to 
deeply analyze the spatial effect of carbon market trading on 
the green transformation of industrial enterprises, in order to 
realize the coordinated governance of the region, and to 
promote the energy conservation and emission reduction of 
industry. 

2. Literature Review 
The operation of carbon markets as a policy tool for 

governments to regulate carbon emissions has implications in 
a number of ways, with Johnstone[1] et al. demonstrating that 
market-based instruments such as environmentally related 
taxes and trading licenses are more likely to induce 
innovation than direct regulations such as technology-based 
standards. Meanwhile, some scholars believe that 
environmental regulations will increase the production costs 
of enterprises to a certain extent[2] , while others believe that 
environmental regulatory policies will incentivize enterprises 
to innovate, thus improving the productivity of enterprises 
and bringing additional benefits[3] , both of which illustrate 
the policy effects of environmental regulations on enterprise 
innovation. Specifically for the carbon trading market, 
domestic scholars have studied the results of the carbon 
trading market from different perspectives. Wang Yong[4] et 
al. study the impact of carbon trading policy on the 
environment and find that the establishment of carbon market 
has the effect of improving carbon emission efficiency. Liu 
Chuanming[5] et al. used the synthetic control method to 
study the emission reduction effect of the policy at the 
national level, and found that the carbon trading policy brings 
obvious carbon emission reduction effect, but the emission 
reduction effect varies in different provinces due to the 
differences in the level of economic development and energy-
saving technology. Lu Min[6] et al. used trend analysis and 
gray prediction model and found that carbon trading 
mechanism can effectively reduce the indirect carbon 
emission intensity of Shanghai's industrial sector, but the 
impact on direct carbon emissions is small. Xue Fei[7] et al. 
confirmed that there are carbon emission reduction effect and 
synergistic emission reduction effect of carbon trading market 
scale, the expansion of carbon trading market scale is 



 

70 

conducive to the reduction of carbon emissions in the pilot 
area, and the carbon trading market scale can reduce sulfur 
dioxide emissions while realizing the carbon emission 
reduction and synergistic emission reduction effect at the 
same time. Ren Xiaosong[8] et al. studied the effect and 
heterogeneity of carbon trading policy on industrial carbon 
productivity, and found that the carbon trading policy 
significantly improved industrial carbon productivity, 
realizing the effect of "reducing carbon and promoting 
economic development", but there are some differences in the 
degree of response to the policy in each pilot region. 

Green technological innovation has been regarded as an 
important means to reduce environmental pollution, and the 
research on green technological innovation can be traced back 
to the 1990s, when C. Fussler[9] and others firstly put forward 
and defined "green innovation", and regarded that green 
technological innovation is a kind of new product and new 
technology, which can bring good economic value to the 
enterprise, the consumer and the society. It is also considered 
that green technology innovation is a new product and 
technology that can bring good economic value to enterprises, 
consumers and society. Measurement of green technology 
innovation efficiency. To measure the efficiency of green 
technological innovation, it is mainly carried out from the 
aspects of evaluation object[10] and evaluation method[13] . 
Another research hotspot is the influencing factors of green 
innovation efficiency. Han Jing[16] et al. believe that foreign 
investment and the restructuring of productive service 
industries are conducive to the improvement of green 
innovation efficiency; Li Jinri[17] et al. believe that the 
degree of urban transportation, government support, the 
number of urban colleges and universities, as well as the 
development of tertiary industry have a positive effect on the 
promotion of green innovation efficiency; however, Yang 
Shuwang[18] et al. believe that there is not a significant 
relationship between the industrial structure and green 
innovation, and they believe that enterprise pollution cost, 
technology market and the development of the industry are 
not important factors. They believe that enterprise pollution 
cost, technology market maturity and market openness can 
promote the efficiency of green innovation. 

To summarize, at present, most of the studies conducted by 
scholars at home and abroad on how the carbon trading 
market affects carbon emission efficiency and the efficiency 
of green technological innovation focus on the measurement 
of the efficiency of green technological innovation as well as 
the influencing factors. However, due to the late opening of 
the domestic carbon market, and because of the selection of 
indicators and different research perspectives, the conclusions 
of the relevant studies are also different. This literature lays a 
good foundation for the subsequent research of this paper, but 
there are still some limitations. First of all, in the 
measurement of green innovation efficiency, it does not take 
into account that green innovation is a process, but is simply 
viewed as a whole. Secondly, in terms of research content, the 
current scholars mostly focus on the direct impact of carbon 
market on enterprises' green technological innovation, and do 
not explore it from the perspective of spatial effect, especially 
under the strategy of regional synergistic development, the 
mutual influence between provinces in China is more and 
more important. Therefore, based on the synthesis of previous 
research, this paper firstly constructs a two-stage value chain 
model of green innovation of Chinese industrial enterprises 
through a two-stage DEA model, and empirically analyzes the 

spatial effect of carbon market on industrial green 
technological innovation based on this using panel data from 
2005 to 2020, and comprehensively examines the incentive 
effect of low-carbon city construction on the efficiency of 
urban green innovation from the spatial perspective. 

3. Model Construction and Variable 
Description 

3.1. Super-efficiency DEA model 
Traditional DEA models are mainly based on CCR or BBC, 

and with the deepening of research, DEA models have been 
optimized and extended, deriving such models as DEA-SBM, 
DEA-RAM, etc. Although these models solve the problems 
of input-output slack variables and non-desired outputs in the 
evaluation of efficiency, they all look at the innovation 
process as a whole and only consider the overall inputs and 
outputs and thus neglect the stages of the innovation process. 
Based on the innovation value chain theory proposed by Qian 
Li[19] et al. in this paper, we constructed a two-stage value 
chain model of green innovation for Chinese industrial 
enterprises, as shown in Figure 1. 

 

 
Figure 1. Two-stage value chain model of green innovation 

in industrial companies 
 

In this study, the super-efficiency DEA model is used to 
calculate the green science and technology R&D efficiency 
and green achievement transformation efficiency of industrial 
enterprises in 30 provinces of China from 2005 to 2020, 
respectively, and the overall green innovation efficiency is 
expressed by the product of science and technology R&D and 
achievement transformation efficiencies, which is borrowed 
from the study of Yu Yongze[20] et al. The formula of the 
super-efficiency DEA model is: 

 
Min θ 

𝑋 𝜆 𝑆 θ𝑋  

∑ 𝑌 𝜆 𝑆 𝑌                 (1) 
𝜆 0, 𝑗 1,2 … , 𝑘 1, 𝑘, … , 𝑛 

𝑆 0, 𝑆 0 
   

where θ denotes the efficiency values of𝑋   and𝑌   denote 
the input and output variables of the jth decision unit, 
respectively.𝑆  and𝑆  denote slack variables, n denotes the 
number of decision units, and λ is a column vector. 

3.2. Double Difference Spatial Durbin Model 
(SDM-DID) 

3.2.1. Parallel trend test  
The double difference model (DID) is valid provided that 

the parallel trend assumption is satisfied. Double differencing 



 

71 

does not require that the experimental and control groups are 
identical, but it does require that the experimental and control 
groups have the same trend over time. Therefore a parallel 
trend test is required before using the double difference model. 
The event study method is used to empirically test the 
dynamic effect of carbon trading policy. The model is as 
follows: 

 
𝐺𝐼𝐸 β ∑ 𝛽 𝑑 𝛽 ∑ 𝑥 𝛾 𝜆 𝜀   (2)   
 
where 𝐺𝐼𝐸   denotes the green innovation efficiency of 

each province in each year, and 𝑥   denotes the control 
variables;𝑑  is the combined dummy variable, the𝑑 𝑑 ∗
𝑑   and𝑑   and𝑑   are all 0-1 variables.𝑑   The assignment is 
when the year is defined as 0 in the year before the 
implementation of the carbon emissions trading policy (2005-
2013) and 1 in the year after the implementation of the carbon 
emissions trading policy (2014-2020).𝑑   Indicates whether 
the province has implemented the carbon emissions trading 
policy or not, and the regions that have implemented it 
(Beijing, Tianjin, Shanghai, Chongqing, Hubei, and 
Guangdong) are assigned the value of 1, while the remaining 
24 provinces are assigned the value of 0.𝛾  denotes time fixed 
effects.𝜆  denotes individual fixed effects, and𝜀 denotes the 
random error term.𝛽   The coefficients of the estimates are 
insignificant if they contain 0, and significant if they do not. 
If the coefficient of the interaction term before policy 
implementation is not significant and the coefficient of the 
interaction term after policy implementation is significant, it 
means that there is no significant difference between the 
control group and the experimental group before policy 
implementation, and the parallel trend test is passed. 

3.3. Double Difference Modeling (DID) 
The double difference model is often used to evaluate the 

effect of policy implementation, the principle of which is to 
divide the research subjects into experimental group and 
control group, and compare the effect before and after policy 
implementation by controlling the systematic difference 
between the two groups. The change before and after the 
control group represents the change brought about by time, 
and the change before and after the experimental group 
represents the change brought about by policy interference 
and time at the same time, and the change brought about by 
policy implementation can be obtained by subtracting the 
change in the experimental group from the change in the 
control group. This paper establishes the following model 
based on the double difference method. The experimental 
group is set to be the six provinces and cities where the policy 
is implemented, and the control group is set to be other 
provinces and cities; at the same time, 2013, the year before 
the implementation of the policy, is used as the base period, 
and the year dummy variable is defined, with the year after 
2013 defined as 1 and the year before 2013 defined as 0. The 
model is constructed as follows: 

 

𝐺𝐼𝐸 β β 𝑑 β 𝑑 β 𝑑 𝛽 𝑥 𝛾 𝜆 𝜀   3  

 
The definitions of the variables in this model are the same 

as in the previous model. 

3.4. Spatial measurement models 
Since different types of spatial econometric models assume 

different spatial transmission mechanisms and economic 
significance, the selection of spatial econometric models is of 
great importance[21] . In order to obtain a better fitting effect, 
a series of spatial econometric model tests are carried out with 
reference to Elhorst[22] et al. before the regression is carried 
out, and then the spatial econometric model suitable for the 
study of this paper is selected. First, the LM test and the robust 
LM test are used to determine the corresponding fit of the 
SEM model, the SAR model, and the SDM model; second, 
the LR test is conducted to further determine whether the 
SDM model degrades into the SAR or SEM model; finally, 
the Hausman test is used to determine whether to select the 
random-effects or the fixed-effects model for the study. The 
testing process followed the path of spatial autocorrelation 
model (SAR), spatial error model (SEM), and spatial Durbin 
model (SDM). To eliminate the effect of heteroskedasticity, 
the non-ratio variables in the model were logarithmized. 

(1) Spatial Autocorrelation Model (SAR) 
The spatial autocorrelation model, also known as the 

spatial lag model (SLM), in which spatial dependence 
between variables leads to spatial correlation, and there is 
unidirectional spatial correlation between regions. The model 
contains a lag term for the spatial dependent variable. The 
following equation: 

 
𝑌 ∝ 𝛾𝑊𝑌 𝑋 β 𝑐 𝑢 𝜀               4  

 
where W denotes the spatial weight matrix and 𝑊𝑌  

denotes the𝑌  the spatial lag term, and𝑐  denotes the time-
fixed, the 𝑢   denotes individual fixed, and 𝜀   denotes the 
random error term. 

(2) Spatial Error Model (SEM) 
In the SEM model, technological innovation spillovers are 

due to random shocks whose spatial effects are mainly 
transmitted through the error term, and the model includes the 
random error autocorrelation term as the spatial error term. 
The following equation: 

 
𝑌 ∝ 𝑋 β 𝑐 𝑢 𝜀                       5  

 
𝑣 ρ𝛾𝑊𝑌 𝜀                            6  

 
where W is a known spatial weight matrix and there is a 

spatial dependence of the perturbation term, if the coefficient 
of the perturbation termρ = 0, then the model can be reduced 
to a general linear regression model. 

(3) Spatial Durbin Model (SDM) 
In the SDM model, both the change in the error term due to 

the spatial lag of the dependent variable and the spatial 
interactions between regions are taken into account, as 
modeled below: 

 
 𝑌 ∝ 𝛾𝑊𝑌 𝑋 β 𝜃𝑊𝑋 𝑐 𝑢 𝜀    7  

 
where W is the known spatial weight matrix, WX denotes 

the effect of explanatory variables from neighboring regions, 
and other variables have the same meaning as above. 

3.5. Double Difference Spatial Durbin Model 
(SDM-DID) 

The non-spatial double-difference method only studies the 
impact of the policy implementation on the local area, 
ignoring the impact of the policy implementation on the 
neighboring areas of the implementation area. By 



 

72 

implementing the carbon trading policy in the pilot province, 
the government promotes green technology innovation in the 
province, thus achieving the effect of reducing carbon 
emissions in the pilot area; at the same time, because the pilot 
city and the surrounding cities are not independent of each 
other, and have extensive economic ties, when the distance 
between the provinces is closer, the closer the ties between 
the provinces are; therefore, when the carbon trading market 
policy is implemented, the pilot city's surrounding cities will 
also be indirectly affected. Therefore, to evaluate the impact 
of carbon trading market policy, it is not only necessary to 
consider the impact on the pilot provinces, but also need to 
consider the spillover effect on the green innovation 
efficiency of the neighboring provinces of the pilot provinces. 
Therefore, in order to study the impact of carbon trading 
market policy on carbon emission efficiency and its spatial 
spillover effect, this paper introduces the spatial weight 
matrix on the basis of the non-spatial double-difference model, 
and constructs the following double-difference spatial model: 

 

𝐺𝐼𝐸 𝛼 α 𝑑 α 𝑊 ∗ 𝑑 𝛼 𝑥 𝛾 𝜆 𝜀    8  

 
where W in Eq. denotes the spatial weight matrix. The 

following two spatial weight matrices are used in this paper. 
The spatial adjacency matrix, constructed based on whether 

or not each province is in a geospatially adjacent position, 
When two provinces are adjacent，𝑤 =1；otherwise，𝑤 =0 

The economic distance matrix, a spatial weighting matrix 
constructed with the economic level of each province, is set 
as follows: 

 

 
 
Of these, the𝐺𝐷𝑃  and𝐺𝐷𝑃  denote the per capita GDP of 

provinces and provinces. 

3.6. Description of data sources and variables 
3.6.1. Explained Variables 

In this paper, the green innovation efficiency of industrial 
enterprises in each province is taken as an explanatory 
variable and is calculated using the super-efficiency DEA 
model. According to the data availability and the consistency 
of statistical caliber, this paper finally selects the panel data 
of 30 provinces and cities in China except Tibet, Hong Kong, 
Macao and Taiwan from 2005 to 2020 to measure the green 
innovation efficiency of industrial enterprises in each 
province and city in China. The specific input-output 
indicators are as follows: 

(1) Green technology research and development phase 
In this paper, the input indicators of science and technology 

R&D stage are selected from three aspects: labor, R&D 
capital and energy consumption, which are measured by the 
full-time equivalent of R&D personnel, the stock of internal 
expenditure of R&D funding, the stock of new product R&D 
funding and the total amount of coal consumption, 
respectively. Among them, the internal expenditure stock of 
R&D funds is calculated by the perpetual inventory method 
with reference to Xiangdong Li[23] et al. The method is as 
follows: 

 

𝐾 , 𝐼 , +(1-θ)* 𝐾 ,                (9) 
 

 
Among them.𝐾 ,   , the𝐾 ,   denote the stock of internal 

expenditure on R&D funding in province i in period t and 
period t-1, respectively, θ is the depreciation rate, and θ is 
taken to be 15%. 𝐼 ,   is the constant price R&D funding 
internal expenditure of province i in period t. In which the 
calculation of the stock of internal expenditure of funding in 
the initial period is referred to Ying Li[24] et al. for the 
following calculation: 

 
𝐾 , 𝐼 , /（𝑔 θ）            (10) 

 
𝑔  denotes the average annual growth rate of internal 

expenditure of the constant price funding in province i, where 
the geometric mean method is used for calculation. And 
taking 2005 as the base period, the R&D deflator price index 
is utilized to deflate the funding category indicators, in which 
referring to the research of Zhu Pingfang[25] et al. the R&D 
deflator price index = 0.45*Fixed Asset Investment Price 
Index + 0.55*Consumer Price Index. Meanwhile, this paper 
adopts the number of professional applications and the 
number of new product development projects to measure the 
output of science and technology R&D stage. 

(2) Green Results Transformation Phase 
In this paper, in addition to the number of professional 

applications and the number of new product development 
projects, the stock of enterprise's introduction, digestion and 
absorption funding is also selected as an input indicator for 
the green results transformation stage, and as above, deflated 
and stocked calculations are carried out. 

For the outputs of the transformation stage of green results, 
this paper adopts the sales revenue of new products, regional 
industrial added value, and regional gross domestic product 
as desired outputs, and selects carbon dioxide emissions, 
sulfur dioxide emissions, general industrial solid waste 
generation, and industrial smoke (dust) emissions as non-
desired outputs. 

3.6.2. Explanatory variables 
Since June 2013, in order to implement the policy of 

gradually establishing a domestic carbon market in the 12th 
Five-Year Plan, the carbon market has been piloted in seven 
provinces and municipalities, namely Beijing, Tianjin, 
Shanghai, Chongqing, Hubei, Guangdong and Shenzhen, and 
in order to unify the scope of the study, this paper combines 
Shenzhen with Guangdong Province. And the period of 2005-
2013 is selected as the non-pilot period and the period of 
2014-2020 as the pilot period; Beijing, Tianjin, Shanghai, 
Hubei, Chongqing, and Guangdong are taken as the 
experimental group, and the remaining 24 provinces and 
municipalities are taken as the control group. 

Variables 𝑑   is defined as 0 when the year before the 
implementation of the carbon emissions trading policy (2005-
2013), and 1 when the year after the implementation of the 
carbon emissions trading policy (2014-2020), and the 
variable𝑑   denotes whether the province has implemented 
the carbon emissions trading policy, and the regions that have 
implemented it (Beijing, Tianjin, Shanghai, Chongqing, 
Hubei, Guangdong) are assigned the value of 1, while the 
remaining 24 provinces are assigned the value of 0. The 
variable𝑑 𝑑 ∗ 𝑑  as explanatory variables to indicate the 
net effect of carbon trading market policies. 



 

73 

3.6.3. Control variables 
This paper takes into account the variability in the level of 

economic development, the degree of openness to the outside 
world, the industrial structure, the strength of government 
support and the population density of the city, as well as the 
fact that these factors may have an impact on the efficiency 
of green innovation, and so the above factors are used as 
control variables. 

(1) Level of economic development (PGDP) 
The level of economic development has a great impact on 

carbon emissions. Rapid economic development accelerates 
the demand for goods and services by individuals and 
enterprises, which leads to an increase in industrial 
production activities and ultimately leads to an increase in 
carbon emissions, while provinces and cities with higher 
carbon emissions have a more urgent need to improve the 
efficiency of green innovation. So this paper refers to the 
research of Zhu Huan[26] et al. and selects the logarithm of 
per capita GDP of each province and city to measure the 
economic development level of each province and city. 

(2) Level of Foreign Investment (FDI) 
Foreign investment will bring competition effect to the 

provinces and cities where foreign investment is introduced. 
In order to seize the market with local enterprises, foreign-
invested enterprises must reduce costs, improve product 
quality, and enhance their innovation ability; at the same time, 
when foreign-invested enterprises enhance their innovation 
ability, local enterprises will enhance their innovation ability 
in order to prevent being eliminated from the market. This 
paper refers to the practice of Liu Haiyun[27] et al. and selects 
the logarithm of the level of foreign direct investment in each 

province and city to measure foreign investment. 
(3) Industrial Structure (IS) 
Compared with other industries, the secondary industry has 

a higher intensity of energy demand and utilization, so the 
higher the proportion of the secondary industry, the higher the 
energy consumption, which will lead to a larger carbon 
emission, and it will be more necessary to improve the 
efficiency of green innovation. In this paper, we refer to Li 
Wenhong[28] et al. who choose the logarithm of the ratio of 
the value added of the secondary industry to the GDP of each 
province and city to measure the industrial structure of each 
province and city. 

(4) Level of government support (GS) 
Local governments are an important support for 

technological innovation activities in provinces and 
municipalities, and government support can increase the 
motivation of innovative enterprises and thus improve their 
innovation efficiency. In this paper, we refer to Xiao 
Liming[29] et al. and select the logarithm of the ratio of local 
government education plus S&T expenditures to the GDP of 
each province and city to measure the strength of government 
support in each province and city. 

(5) Urban Population Density (UPD) 
According to the IPAT model (Environmental Impact = 

Population*Affluence*Technology) proposed by Ehrlich and 
Holdren[30] , regional human activities also have an impact 
on environmental change, which affects the regional green 
innovation efficiency, and therefore urban population density 
is included in the model as a control variable. . 

The specific variables are described in the table below. 

 
Table 1. Description of variables 

control variable notation Description of calculations 

Level of economic development PGDP 
GDP of provinces and municipalities / resident population of 

provinces and municipalities; logarithmic scale 

Level of foreign investment FDI 
FDI in provinces and municipalities/GDP in provinces and 

municipalities; logarithmic scale 

industrial structure IS 
Value added of secondary industry in provinces and 

municipalities/GDP in provinces and municipalities; logarithmic 
scale 

Government support GS 
Local government expenditure on education plus science and 

technology/GDP of provinces and municipalities; logarithmic scale 

Urban population density UPD 
Number of resident population in provinces and 

municipalities/area of provinces and municipalities; logarithmic 
scale 

 
3.6.4. Data sources 

In this paper, panel data of 30 provinces and cities in China 
from 2005 to 2020 are selected, and Tibet, Hong Kong, 
Macao and Taiwan are not included in the analyzed samples 
due to their serious missing data. The raw data of this paper 
come from China Statistical Yearbook, China Environmental 
Statistical Yearbook, China Science and Technology 
Statistical Yearbook, China Energy Statistical Yearbook, 
China Trade and Foreign Economy Statistical Yearbook, as 
well as statistical yearbooks of provinces and cities. 
According to the two-stage DEA model in the previous 
section, the green innovation efficiency of industrial 

enterprises in 30 provinces in China from 2005 to 2020 is 
calculated as shown in the following table: 

According to the table we can know that the provinces with 
high green innovation efficiency are Guangdong, Beijing and 
Shanghai, while the provinces with low green innovation 
efficiency are Ningxia, Inner Mongolia and Gansu, and the 
ones in the middle ranking are mostly the provinces and cities 
located in the center. Because the innovation efficiency is 
closely related to the economic development and other 
conditions of the provinces and cities, the cities located in the 
coastal area will be higher than the innovation efficiency in 
the inland area, and at the same time, this also reflects a 
certain aggregation of high innovation efficiency. 

 
 
 
 



 

74 

Table 2. Green innovation efficiency of industrial enterprises, 2005-2020 
𝑮𝑰𝑬𝒊𝒕 2005 2006 2007 2008 2009 2010 2011 2012 
Anhui 0.801 0.809 0.818 0.824 0.835 0.855 0.859 0.864 

Beijing, 0.883 0.888 0.918 0.914 0.891 0.897 0.894 0.893 
Fujian 0.895 0.897 0.907 0.910 0.913 0.919 0.914 0.911 
Gansu 0.782 0.803 0.816 0.814 0.811 0.831 0.831 0.832 

Guangdong 0.950 0.958 0.956 0.955 0.956 0.964 0.966 0.963 
Guangxi 0.829 0.839 0.849 0.848 0.845 0.855 0.858 0.854 
Guizhou 0.766 0.775 0.784 0.798 0.798 0.808 0.810 0.816 

Hainan Island 1.007 1.012 1.010 1.001 1.000 1.004 1.001 0.962 
anhui 0.839 0.849 0.860 0.869 0.861 0.867 0.857 0.858 

He'nan 0.852 0.866 0.879 0.892 0.888 0.897 0.893 0.895 
Heilongjiang 0.842 0.847 0.855 0.863 0.854 0.857 0.861 0.865 

Hubei 0.822 0.826 0.841 0.849 0.857 0.874 0.875 0.885 
Hunan 0.841 0.853 0.863 0.867 0.869 0.880 0.879 0.882 
Jilin 0.759 0.769 0.784 0.798 0.860 0.816 0.822 0.829 

Jiangsu 0.876 0.885 0.894 0.900 0.901 0.907 0.899 0.904 
Jiangxi 0.835 0.851 0.861 0.870 0.873 0.895 0.896 0.901 

Liaoning 0.786 0.797 0.806 0.821 0.823 0.822 0.823 0.826 
Inner Mongolia 0.781 0.794 0.801 0.809 0.816 0.816 0.810 0.816 

Ningxia 0.804 0.799 0.787 0.791 0.790 0.810 0.790 0.796 
Qinghai 1.011 1.005 0.906 0.883 0.862 0.838 0.824 0.835 

Shandong 0.861 0.875 0.882 0.885 0.889 0.890 0.878 0.881 
Shanxi 0.787 0.793 0.809 0.819 0.812 0.828 0.835 0.834 
Shaanxi 0.854 0.864 0.868 0.876 0.877 0.889 0.886 0.894 

Shanghai 0.906 0.913 0.918 0.914 0.904 0.915 0.908 0.904 
Sichuan 0.809 0.825 0.838 0.852 0.860 0.870 0.875 0.882 
Tianjin 0.818 0.833 0.827 0.835 0.832 0.840 0.835 0.843 

Xinjiang 0.937 0.926 0.893 0.888 0.851 0.880 0.876 0.870 
Yunnan 0.857 0.865 0.873 0.874 0.867 0.907 0.862 0.871 
Zhejiang 0.875 0.881 0.892 0.894 0.890 0.902 0.895 0.898 

Chongqing 0.834 0.844 0.849 0.851 0.856 0.869 0.874 0.880 
𝑮𝑰𝑬𝒊𝒕 2013 2014 2015 2016 2017 2018 2019 2020 
Anhui 0.869 0.870 0.869 0.872 0.885 0.891 0.890 0.909 

Beijing, 0.903 0.915 0.926 0.939 0.967 0.972 0.998 1.022 
Fujian 0.911 0.915 0.919 0.914 0.929 0.936 0.935 0.965 
Gansu 0.832 0.829 0.818 0.815 0.820 0.828 0.835 0.847 

Guangdong 0.965 0.971 1.001 1.029 0.997 1.000 0.999 1.008 
Guangxi 0.854 0.855 0.864 0.872 0.880 0.884 0.877 0.886 
Guizhou 0.827 0.838 0.852 0.859 0.863 0.870 0.881 0.890 
Hainan 0.929 0.931 0.941 0.985 0.975 0.995 1.000 1.023 
anhui 0.851 0.851 0.856 0.870 0.876 0.874 0.880 0.895 

He'nan 0.898 0.900 0.908 0.924 0.944 0.949 0.955 0.970 
Heilongjiang 0.863 0.858 0.844 0.839 0.843 0.854 0.857 0.862 

Hubei 0.891 0.896 0.904 0.919 0.931 0.943 0.945 0.945 
Hunan 0.885 0.886 0.892 0.899 0.899 0.903 0.916 0.931 
Jilin 0.829 0.838 0.841 0.851 0.854 0.866 0.855 0.865 

Jiangsu 0.903 0.903 0.911 0.917 0.928 0.942 0.954 0.982 
Jiangxi 0.905 0.904 0.903 0.903 0.908 0.908 0.912 0.998 

Liaoning 0.830 0.824 0.822 0.825 0.829 0.835 0.840 0.852 
Inner Mongolia 0.816 0.817 0.821 0.827 0.841 0.849 0.855 0.861 

Ningxia 0.798 0.788 0.785 0.784 0.797 0.798 0.798 0.803 
Qinghai 0.858 0.852 0.838 0.870 0.887 1.000 0.961 1.001 

Shandong 0.882 0.878 0.882 0.893 0.901 0.901 0.905 0.920 
Shanxi 0.829 0.822 0.810 0.814 0.844 0.856 0.863 0.868 
Shaanxi 0.896 0.897 0.892 0.898 0.918 0.949 0.964 0.906 

Shanghai 0.899 0.899 0.899 0.917 0.958 0.966 0.971 0.985 
Sichuan 0.888 0.887 0.893 0.901 0.906 0.913 0.922 0.929 
Tianjin 0.866 0.861 0.869 0.887 0.873 0.883 0.891 0.903 

Xinjiang 0.869 0.885 0.862 0.857 0.884 0.926 0.970 0.954 
Yunnan 0.879 0.882 0.883 0.879 0.888 0.898 0.898 0.904 
Zhejiang 0.904 0.908 0.924 0.947 0.941 0.951 0.965 1.002 

Chongqing 0.888 0.892 0.896 0.908 0.905 0.902 0.908 0.929 
 
 
 
 
 



 

75 

 

4. Empirical Analysis and Results 

4.1. Parallel trend test 
The prerequisite for the realization of the double-difference 

method is to pass the parallel trend test, and the following 
figure shows the parallel trend dynamic effect test plot of this 
paper, where the vertical dashed line indicates the𝑑  the 95% 
confidence interval of the regression coefficient, in order to 
avoid multicollinearity, this paper sets the year before the 
implementation of the policy, which is 2013, as the base 
period. According to the figure, we can see that in the first 
five years of policy implementation, the 95% confidence 

intervals of the regression coefficients contain 0, indicating 
that the regression coefficients are not significant; while in the 
first year after the implementation of the policy, the 95% 
confidence intervals of the regression coefficients cross the 
horizontal line, indicating that the regression coefficients are 
significant, and the coefficients change from negative to 
positive, which indicates that the implementation of the 
policy has a positive effect on the promotion of the 
explanatory variables and has been influential for several 
years after the implementation. has been influential for 
several years afterward. 

 

 
Figure 2. Plot of the test for the dynamic effect of parallel trends 

 

4.2. Spatial correlation analysis  
Airflow, water flow and industrial transfer affect carbon 

emissions in neighboring provinces, which indirectly affects 
the efficiency of green innovation in neighboring provinces. 
Therefore, in order to scientifically verify the spatial 
correlation of carbon emissions, this paper constructs a 0-1 
proximity matrix based on geographic proximity, and 

measures the Moran's I value of carbon emissions of 30 
provinces in China from 2005 to 2020. The results are shown 
in the following table, the Moran's indices of carbon 
emissions are all positive and pass the significance level. This 
indicates that there is a certain spatial clustering effect of 
carbon emissions among Chinese provinces, which provides 
a scientific basis for borrowing spatial measurement as an 
analytical tool. 

 
Table 3. Moran's I index of green innovation efficiency, 2005-2020 

year Moran's I E(I) Sd(I) z p-value 
2005 0.148 -0.034 0.110 1.657 0.049 
2006 0.152 -0.034 0.110 1.697 0.045 
2007 0.177 -0.034 0.110 1.925 0.027 
2008 0.173 -0.034 0.110 1.889 0.029 
2009 0.172 -0.034 0.108 1.910 0.028 
2010 0.223 -0.034 0.109 2.357 0.009 
2011 0.247 -0.034 0.109 2.593 0.005 
2012 0.290 -0.034 0.111 2.921 0.002 
2013 0.357 -0.034 0.111 3.516 0.000 
2014 0.339 -0.034 0.111 3.361 0.000 
2015 0.354 -0.034 0.110 3.523 0.000 
2016 0.292 -0.034 0.110 2.975 0.001 
2017 0.287 -0.034 0.112 2.869 0.002 
2018 0.170 -0.034 0.112 1.819 0.034 
2019 0.166 -0.034 0.113 1.776 0.038 
2020 0.350 -0.034 0.113 3.106 0.000 



 

76 

 

4.3. Model Testing 
In order to scientifically demonstrate the relationship 

between variables, this paper introduces a variety of spatial 
econometric models to be empirically analyzed. Referring to 
the research conclusion of Anselin[31] , this paper 
successively conducts LM-Error test, LM-Lag test, LM-Error 
robust test, LM-Lag robust test, Hausman test and LR test, 
and the test results are shown in the following table, which 
can be seen that the values of the LM-Error test and LM-Lag 
test pass the 1% significance level test, indicating that the 
SDM model should be considered in a limited way; the 
Hausman test also passes the 1% significance level test, using 
fixed effects; while at the same time, the SDM model is used. 
significance level test, indicating that the SDM model should 
be considered in a limited way; the Hausman test also passes 
the 1% significance level test, using fixed effects; and also 
passes the LR test, with both the time and spatial fixed effects 
significant at 1%. So this paper uses a spatial panel Durbin 

model with time and space double fixed. 
 

Table 4. Model Tests 
Test Methods statistical value P-value 

LM-Lag 80.042 0.000 
LM-Lag(robust) 115.035 0.000 

LM-Error 53.691 0.000 
LM-Error(robust) 88.684 0.000 

Hausman 84.20 0.000 
LR-spatial fixed 

effects 
43.40 0.000 

LR-time-period fixed 
effects 

509.16 0.000 

4.4. Empirical results of spatial econometric 
modeling 

Based on the above process, the empirical results of this 
paper are shown in the following table. 

 
Table 5. Empirical results of DID and SDID models 

variant 
DID SDID 

Coef. Std. Err t Coef. Std. Err t 
𝑑  0.028 0.005 4.87*** 0.026 0.005 6.11*** 

PGDP 0.027 0.024 2.09** 0.043 0.128 3.38*** 
FDI 0.008 0.003 2.76*** 0.006 0.002 3.40*** 
IS 0.079 0.025 3.12*** 0.060 0.016 3.77*** 
GS 0.009 0.012 0.71 -0.010 0.007 -1.34** 

UPD -0.001 0.006 -0.54 -0.005 0.003 -1.670** 
rho - - - 0.269 0.062 4.33*** 

Sigma2_e - - - 0.002 0.001 15.33*** 
Note: ***, ** and * sub-tables indicate significance levels of 1%, 5% and 10%. 

 
From the above table, it can be seen that the explanatory 

variable in column DID 𝑑   denotes the effect of carbon 
trading policy on green innovation efficiency, and its 
coefficient is 0.028 and is significant at 1% significance level, 
indicating that carbon trading policy has a significant positive 
promotion effect on green innovation efficiency in pilot cities. 
This is because the implementation of carbon trading policy 
in the pilot city promotes the carbon reduction behavior of 
industrial enterprises, and in order to achieve the carbon 
reduction target further motivates the green innovation 
efficiency of industrial enterprises. At the same time, the 
coefficients of the economic development level (PGDP), 
foreign investment level (FDI), and industrial structure (IS) 
passed the significance level test of 5%, with the coefficients 
of 0.027, 0.008, and 0.079, respectively, which indicates that 
these three control variables have a positive facilitating effect 
on green innovation efficiency. The level of economic 
development has a positive promotional effect on the green 
innovation efficiency of enterprises because regions with 
higher levels of economic development will have more 
sufficient funds to develop green production technology, and 
at the same time, it can also reduce carbon emissions while 
improving the efficiency of green innovation, forming a 
virtuous circle. The level of foreign investment promotes the 
level of green innovation because foreign-invested 
enterprises will have a competitive effect with local 
enterprises, foreign-invested enterprises will increase 
innovation to compete with local enterprises, and the same 
local enterprises will also increase innovation to ensure that 
they will not be eliminated, thus promoting the rise of the 

overall innovation capacity of enterprises in the region. The 
calculation of industrial structure in this paper is the ratio of 
the value added of the secondary industry to the GDP, and the 
secondary industry is mostly a high-energy-consuming 
industry, with more serious carbon emissions, so the rising 
industrial structure is even more necessary to improve the 
emergence of green production technology. 

The explanatory variable in the SDID column 𝑑   The 
coefficient of the SDID column is 0.026 and is significant at 
1% significance level, which indicates that there is a positive 
spatial correlation effect of green innovation efficiency 
between regions, and this result is also in line with the results 
of the Moran index test. The geographic conditions of 
neighboring regions are similar, which means that the 
resource conditions are also similar, in addition, the economic 
trade and technology circulation in neighboring regions are 
more convenient and fast, and the carbon trading policy 
provides a platform for carbon trading between enterprises, 
which promotes the improvement of technological innovation 
efficiency even more. 

4.5. Spatial spillover effects 
Based on the above, it is known that there may be spillover 

effects due to the carbon trading policy on the green 
innovation efficiency of neighboring provinces. Therefore, 
drawing on the research results of LeSage[32] , partial 
differential equations are used to decompose the spatial 
effects into direct, indirect and total effects, of which the 
indirect effect is the spatial spillover effect, and the specific 
calculation results are shown in the following table. 



 

77 

Table 6. Decomposition of effects for the double 
difference spatial Durbin model 

variant LR_Direct LR_Indirect LR_Total 

𝑑  
0.028*** 

(6.35) 
0.032** 
(2.42) 

0.061*** 
(4.10) 

PGDP 
0.042** 
(3.32) 

-0.018 
(-1.07) 

0.023* 
(1.13) 

FDI 
0.006** 
(3.37) 

-0.001 
(-0.08) 

0.005* 
(1.07) 

IS 
0.061*** 

(3.55) 
-0.011 
(-0.27) 

0.049 
(0.93) 

GS 
-0.007 
(-0.95) 

0.049*** 
(3.75) 

0.041** 
(2.71) 

UPD 
-0.006* 
(-1.75) 

-0.011 
(-1.28) 

-0.016* 
(-1.63) 

Note: Numbers in parentheses are t-statistics, and ***, **, 
and * sub-tables indicate significance levels of 1%, 5%, and 
10%. 

 
From the table, we can see that the direct effect of carbon 

trading policy is 0.028, the spatial spillover effect is 0.032, 
and the total effect is 0.061, and all of them pass the 
significance test of 5%, which indicates that the carbon 
trading policy has a positive promotional effect on the 
innovation efficiency of the pilot cities, and there is also a 
positive spillover benefit, which not only improves the 
innovation efficiency of the pilot cities, but also improves the 
innovation efficiency of the neighboring It can not only 
improve the innovation efficiency of pilot cities, but also 

improve the innovation efficiency of neighboring provinces 
and cities. On the one hand, it is more convenient to circulate 
technology between neighboring cities, and under the 
pressure of green transformation, the policy dividend of the 
pilot city will drive the green innovation of the nearby cities, 
and on the other hand, the leading role of the pilot city will 
lead to the competition effect between the neighboring 
provinces and cities, so that there will be a spillover effect of 
the carbon trading policy on the green innovation efficiency 
of the pilot city. In addition, government support (GS) also 
has a positive spatial spillover effect, and government 
expenditure on education and science and technology 
positively affects innovation efficiency, while the existence of 
innovation city competition pressure from neighboring city 
governments helps neighboring cities to increase science and 
innovation expenditures and improve the level of innovation. 
And several other control variables do not pass the 
significance test, indicating that their spatial spillover effects 
are not formed. 

4.6. Robustness Tests 
4.6.1. Robustness test of propensity score matched 

double difference model (PSM-DID) 
In order to ensure the robustness of the empirical results, 

this paper adopts the propensity score matched double 
difference model (PSM-DID) to test the impact of carbon 
trading market. In this paper, the original panel data are 
matched and finally aggregated into the equilibrium panel 
data. The equilibrium test is conducted after the matching is 
completed, and the results are shown in the table below: 

 
Table 7. Equilibrium test results 

variant 
Unmatched Mean 

%bias 
%reduct 

|bias| 
T-test 

Matched Treated Control t P-value 

PGDP 
U 11.341 10.355 198.6 

98.9 
11.17 0.00 

M 11.341 11.053 2.3 1.62 0.083 

FDI 
U -3.927 -4.298 40.1 

902 
2.28 0.023 

M -3.927 -4.071 -3.9 1.95 0.085 

IS 
U 1.074 0.852 79.4 

78.7 
-6.66 0.00 

M 1.074 0.895 6.9 0.88 0.381 

GS 
U 1.473 2.037 53.6 

93.0 
1.91 0.062 

M 1.473 2.012 3.7 1.058 0.318 

UPD 
U 7.869 7.795 75.2 

98.9 
3.277 0.003 

M 7.869 7.862 -0.2 0.663 0.542 
 

According to the table, we can see that the %bias value of 
the control variables after matching is less than 10% and is 
significantly smaller than the value before unmatched, which 
indicates that there is no obvious gap between the two groups 

and the balance assumption is satisfied. So the matched data 
were subjected to spatial double difference experiments, and 
the results are shown in the following table: 

 
Table 8. Empirical results for propensity score matched double difference 

variant 
DID PSM-DID 

Coef. Std. Err t Coef. Std. Err t 
𝑑  0.028 0.005 4.87*** 0.535 0.005 7.62*** 

PGDP 0.027 0.024 2.09** 0.152 0.528 3.78*** 
FDI 0.008 0.003 2.76*** 0.521 0.072 6.85*** 
IS 0.079 0.025 3.12*** 0.236 0.028 5.28*** 
GS 0.009 0.012 0.71 -0.523 0.547 -2.12** 

UPD -0.001 0.006 -0.54 -0.105 0.054 -1.895** 
rho - - - 0.354 0.163 5.84*** 

Sigma2_e - - - 0.562 0.001 16.54*** 
 

It can be known from the above table that the carbon 
trading market after matching has a more obvious positive 

promotion effect on innovation efficiency, and the effects of 
other control variables on innovation efficiency are not 



 

78 

significantly different from those before matching. 

4.6.2. Robustness tests for spatial spillovers 
In order to ensure the robustness of the empirical results, 

this paper adopts the method of replacing the spatial weight 
matrix to carry out the robustness test, replacing the spatial 
adjacency matrix with the spatial distance squared matrix to 
carry out the regression analysis, and obtaining the results in 
the following table: 

 
Table 9. Robustness test results 

variant LR_Direct LR_Indirect LR_Total 

𝑑  
0.029*** 

(4.27) 
0.005** 
(1.85) 

0.035*** 
(4.19) 

PGDP 
0.152*** 

(4.32) 
-0.028 
(-0.98) 

0.057* 
(1.08) 

FDI 
0.0246** 

(3.51) 
-0.021 
(-0.06) 

0.015* 
(1.17) 

IS 
0.074*** 

(3.88) 
-0.024 
(-0.35) 

0.047 
(0.87) 

GS 
-0.014 
(-0.82) 

0.045*** 
(3.69) 

0.078*** 
(3.58) 

UPD 
-0.014* 
(-1.85) 

-0.021 
(-1.11) 

-0.048* 
(-1.25) 

 
The comparison shows that the size, direction and 

significance level of the direct effect, indirect effect and total 
effect coefficients of the explanatory variables have not 
changed significantly; although the significance of some of 
the variables has changed, the direction of the coefficients is 
still the same. Overall, the results obtained after replacing the 
matrix are basically the same as those before replacement, 
indicating that the results of this paper are robust from the 
perspective of considering the spatial matrix. 

5. Conclusions and Recommendations 

5.1. Conclusion 
Based on the two-stage DEA model, this paper constructs 

a two-stage value chain model of green innovation of Chinese 
industrial enterprises, measures the green innovation 
efficiency of industrial enterprises in each province and city 
in China, and explores the impact and spatial spillover effect 
of the carbon trading market policy on the green innovation 
efficiency of industrial enterprises in each province and city 
through double-difference method and spatial econometric 
modeling, and the main conclusions and contributions are as 
follows: 

(1) The green innovation activities are decomposed into 
two stages: green science and technology research and 
development and green achievement transformation, and the 
green innovation efficiency of industrial enterprises in 30 
provinces and cities in China is scientifically and reasonably 
measured. Among them, coastal cities such as Guangdong, 
Beijing and Shanghai have higher innovation efficiency, 
while inland cities such as Ningxia, Inner Mongolia and 
Gansu have lower efficiency. 

(2) Carbon trading market policy significantly affects the 
green innovation efficiency of industrial enterprises, while the 
level of economic development, foreign investment level, and 
industrial structure of each province and city also have a 
positive effect on green innovation efficiency. 

(3) There is a spatial spillover effect of the carbon trading 
market policy, which not only has a significant incentive 
effect on the green innovation efficiency of the pilot cities, but 

also has a positive promotion effect on the green innovation 
efficiency of neighboring cities. The robustness test shows 
that the spatial spillover effect still exists even if the spatial 
weight matrix is replaced. At the same time, the support of 
provincial and municipal governments significantly affects 
the spillover effect of carbon market policies. 

5.2. Recommendations 
(1) Emphasizing the spillover effects of carbon trading 

policies to improve the efficiency of green innovation 
Carbon trading policy to improve the efficiency of 

provincial green innovation spillover effect is significant, to 
strengthen communication and cooperation between 
neighboring provinces and regions, across the provinces and 
regions to build a regular, all-round, multi-disciplinary 
technological innovation mechanism, regular meetings and 
communication system, effective use of the synergistic effect 
of the provincial policy, the pilot city of the carbon trading 
policy to play a demonstration leading role. 

(2) Optimization of the regional industrial structure and 
increased investment 

Strengthen the supply of low-carbon technologies, enhance 
the top-level design of the government's science and 
technology innovation system, promote the science and 
technology innovation-driven strategy, improve the incentive 
mechanism for science and technology innovation, and 
promote the upgrading of the industrial structure and 
optimization of the energy structure. Gradually increase the 
proportion of tertiary industry and accelerate the transition 
from secondary to tertiary industry. At the same time, increase 
foreign investment and raise the level of government support. 
Focus on building a platform for scientific and technological 
innovation and research and development, relying on 
scientific and technological innovation power to enhance the 
level of green innovation; adhere to the market orientation, 
optimize the circulation of innovation factors, highlight the 
status of the main body of enterprise innovation, optimize the 
allocation of resources, and improve the conversion rate of 
scientific and technological achievements, and make full use 
of the promotional effect of upgrading the industrial structure 
and foreign investment in the process of improving the 
efficiency of green innovation by the carbon trading policy. 

(3) Rationalizing Carbon Trading Policies and 
Synergistically Enhancing Green Innovation Efficiency 

Rationally allocate the total amount of carbon trading 
among provinces and cities according to the process of 
emission reduction, and design a carbon quota allocation 
system in line with the industrial and energy structure. At the 
same time, it will improve the synergistic governance system 
among cities, actively explore the green innovation policy of 
regional integration and synergy with pilot cities to drive the 
surrounding cities, establish a provincial carbon emission 
monitoring mechanism, adjust the strength of scientific and 
technological innovation policies in a timely manner, and 
promote the upgrading and transformation of the regional 
economic structure, energy efficiency and industrial structure, 
so as to realize the enhancement of the efficiency of green 
innovation. 

References 
[1] Johnstone N, Haščič I, Kalamova M. Environmental policy 

design characteristics and technological innovation: Evidence 
from patent data[J]. 2010. 



 

79 

[2] Kneller R , Manderson E .Environmental regulations and 
innovation activity in UK manufacturing industries [J]. 
Resource & Energy Economics, 2012, 34(2):211-235. 

[3] Porter M E , Linde C V D .Toward a New Conception of the 
Environment-Competitiveness Relationship[J].Journal of 
Economic Perspectives, 1995, 9(4). 97-118.  

[4] WANG Yong,ZHAO Han. Impact of China's carbon trading 
market launch on regional carbon emission efficiency[J]. China 
Population-Resources and Environment,2019,29(01):50-58. 

[5] LIU Chuanming, SUN Zhe, ZHANG Jin. Research on the 
carbon emission reduction policy effect of China's carbon 
emissions trading pilot[J]. China Population-Resources and 
Environment, 2019,29(11):49-58. 

[6] LU Min, CANG Yuquan, LI Yanyan. Impact of carbon trading 
mechanism on industrial carbon emission intensity and 
competitiveness in Shanghai[J]. Technical Economy, 2018, 
37(07):114-120. 

[7] Xue Fei, Zhou Minliang. Research on the emission reduction 
effect of China's carbon trading market size[J]. East China 
Economic Management,2021,35(06):11-21. 

[8] REN Xiaosong, MA Xi, LIU Yujia, ZHAO Guohao. Impact of 
carbon trading policy on industrial carbon productivity and 
transmission mechanism[J]. China Environmental Science, 
2021, 41(11): 5427-5437. 

[9] Fussler C, James P . Driving eco-innovation: a breakthrough 
discipline for innovation and sustainability[J]. pitman pub, 
1996. 

[10] Nasierowski W , Arcelus F J . On the efficiency of national 
innovation systems[J]. Socio-Economic Planning Sciences, 
2003, 37. 

[11] Cao Xia, Yu Juan. Research on regional innovation efficiency 
in China under the perspective of green low-carbon[J]. China 
Population-Resources and Environment,2015,25(05):10-19. 

[12] JINRI LI,CHAO LI,ZEYU LI. Evaluation of urban green 
innovation efficiency and analysis of its influencing factors[J]. 
Statistics and Decision Making,2017(20):116-120. 

[13] ZHANG Feng,REN Shijia,YIN Xiuqing. Green technology 
innovation efficiency of high-tech industries and its scale 
quality threshold effect[J]. Science and Technology Progress 
and Countermeasures,2020,37(07):59-68. 

[14] WANG Hailong, LIAN Xiaoyu, LIN Deming. Empirical 
analysis of the impact of green technology innovation 
efficiency on regional green growth performance[J]. Science 
and Science and Technology Management,2016,37(06):80-87. 

[15] Lv Yanwei, Xie Yanxiang, Lou Xianjun. Study on the spatial 
and temporal jump and convergence trend of green innovation 
efficiency in Chinese regions[J]. Research on Quantitative and 
Technical Economics,2020,37(05):78-97. 

[16] Han Jing. Research on regional green innovation efficiency in 
China[J]. Research on Financial Issues,2012(11):130-137. 

[17] JINRI LI, ZEYU LI, CHAO LI. An empirical study of urban 
green innovation efficiency--evidence from city clusters in the 

middle reaches of the Yangtze River[J]. Journal of Jiangxi 
University of Finance and Economics,2016(06):3-16. 

[18] YANG Shu-wang, WU Ting, LI Zibo. Research on the spatio-
temporal differentiation and influencing factors of green 
innovation efficiency in the Yangtze River Economic Belt[J]. 
Macroeconomic Research,2018(06):107-117+132. 

[19] QIAN Li, XIAO Renqiao, CHEN Zhongwei. Research on 
green technology innovation efficiency of industrial enterprises 
in China and its regional differences--Based on common 
frontier theory and DEA model [J]. Economic Theory and 
Economic Management, 2015 (01): 26-43. 

[20] Yu Yongze. Research on technological innovation efficiency 
of China's high-tech industry and its influencing factors--A 
two-stage analysis based on the perspective of value chain[J]. 
Economic Science,2009(04):62-74. 

[21] Huang G, Zhang J, Yu J, et al. Impact of transportation 
infrastructure on industrial pollution in Chinese cities: a spatial 
econometric analysis [J]. Energy Economics, 2020, 92: 104973. 

[22] Elhorst J P. Matlab software for spatial panels[J]. International 
Regional Science Review, 2014, 37(3): 389-405. 

[23] LI Xiangdong,LI Nan,BAI Junhong,XIE Zhongqiu. Analysis 
of R&D innovation efficiency in high-tech industries[J]. China 
Soft Science,2011(02):52-61. 

[24] Li Ying. Measurement and spatial characterization of R&D 
capital stock in Chinese provinces[J]. Soft Science, 2019, 
33(07): 21-26+33. 

[25] ZHU Pingfang,XU Weimin. The impact of government's 
science and technology incentive policy on R&D investment 
and its patent output of large and medium-sized industrial 
enterprises--an empirical study in Shanghai[J]. Economic 
Research,2003(06):45-53+94. 

[26] ZHU Huan,ZHENG Jie,ZHAO Qiuyun et al. Economic growth, 
energy structure transition and carbon dioxide emission - an 
empirical analysis based on panel data[J]. Economic and 
Management Research,2020,41(11):19-34. 

[27] LIU Haiyun,YAO Weiwei. Two-way FDI, environmental 
regulation and green innovation efficiency[J]. Ecological 
Economy, 2022, 38(10): 131-138+157. 

[28] Li W.H., Cao W.L. FDI, environmental regulation and regional 
green innovation efficiency[J]. Statistics and Decision Making, 
2020, 36(19): 118-122. 

[29] XIAO Liming, GAO Junfeng, LIU Shuai. Changing trends of 
green technology innovation efficiency in China's regions 
based on spatial gradient - an empirical analysis of inter-
provincial panel data[J]. Soft Science,2017,31(09):63-68. 

[30] Ehrlich P R, Holdren J P. Impact of Population Growth: 
complicacy concerning this component of man's predicament 
is unjustified and counterproductive[J]. Science, 1971, 171 
(3977): 1212-1217. 

[31] Anselin L, Bera A K. Introduction to spatial econometrics[J]. 
Handbook of applied economic statistics, 1998, 237(5). 

[32] LeSage J, Pace R K. Introduction to spatial econometrics [M]. 
Chapman and Hall/CRC, 2009.`