How does the use of robots promote carbon reduction?——Spatial correlation and heterogeneity analysis

doi:10.21203/rs.3.rs-2717744/v1

Download PDF

Research Article

How does the use of robots promote carbon reduction?——Spatial correlation and heterogeneity analysis

https://doi.org/10.21203/rs.3.rs-2717744/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 18 Oct, 2023

Read the published version in Environmental Science and Pollution Research →

You are reading this latest preprint version

Along with the continuous improvement of industrial intelligence, the use of robots (UR) is widely used in various aspects of production and life, which is playing an essential role in achieving carbon reduction targets. This paper explores the impact of UR on carbon emissions and analyzes the mechanism by taking 30 provinces in China from 2006 to 2019 as research objects. It found that UR can significantly reduce carbon emissions. However, the degree of marketization plays a masking effect, which limits UR's carbon reduction effect to some extent. Furthermore, the carbon reduction effect of UR is stronger in provinces with lower carbon emissions. Finally, UR has a significant spatial spillover effect on neighboring regions. The improvement of UR level will positively affect the region's and surrounding areas' carbon emission reduction. The relevant findings provide empirical support for further deepening the policy implementation of UR-assisted carbon emission reduction.

the use of robots

carbon reduction

spillover effects

heterogeneity

marketization

masking effect

The report of the 20th National Congress of the CPC proposed that "actively yet prudently promote two-carbon target, based on China's resource endowment, adhere to the principle of first stand before break, and implement carbon peaking action in a planned and sequential way ". This provides a fundamental basis for promoting carbon peak and carbon neutralization and is greatly significant for building a socialist modern country in an all-round way. Over the past 10 years, China's economy has reportedly grown at an average annual rate of 6.6%, energy consumption has increased at an average annual rate of 3%, and energy consumption intensity has decreased by 26.4%. In the face of the strategic tasks of realizing the fundamental improvement of the ecological environment and the construction of ecological civilization of carbon peak carbon neutrality, We need to grasp the new development stage, implement the new development concept, build a new development pattern of strategic height, looking for a breakthrough in carbon emission reduction seize the "bull nose" of green development, coordinate the work related to carbon peak carbon neutrality and ecological environment protection, and accelerate the substantial new progress of carbon peak carbon neutrality.

With the development of artificial intelligence, the number and scope of robots used in Chinese enterprises are increasing. According to the relevant report of the International Federation of Robots (IFR) in 2017, the density of industrial robots in China was 97 sets/10000 people, which exceeded the global average for the first time. In 2021, the density of robots in China will exceed 322 sets/10000 people, ranking the top five in the world. In the context of the first negative growth of China's population and the gradual disappearance of the "population dividend" of economic development, artificial intelligence has a significant impact on economic development, and the widespread use of robots will help to promote the arrival of the "robot dividend" era. At the same time, the application of robots is also of great significance for improving the natural environment and improving the sustainable development ability of social economy. Robots are widely used in the manufacturing industry, which not only improves the production efficiency, but also further improves the sustainable development ability of the industry.

Under the dual-carbon goal, studying whether the use of robots can effectively reduce carbon emissions can not only help governments formulate scientific and effective carbon emission reduction policies, but also improve the sustainable development ability of social and economic life of countries in the digital economy era. Therefore, this paper analyzes the driving effects and impact mechanism of UR on carbon emissions from empirical analysis. Firstly, this paper constructs a theoretical model related to the use of robots and carbon emissions. Secondly, the paper explores the impact of robot use on carbon emissions through fixed effect model. The spatial spillover effect of UR on carbon emissions is further studied through the Spatial Dubin Model (SDM). Finally, with the help of the intermediary effect model, this paper studies the specific mechanism of UR on carbon emissions from the perspective of marketization.

The rest of this paper is organized as follows. Part 2 introduces the literature review. Part 3 presents the empirical model setup and data sources. Part 4 presents the empirical results and discussion. The last part summarizes the full paper and makes corresponding policy recommendations.

2.1 Studies on driving factors of carbon emission reduction

(1) Research perspective: ①Analyze the driving factors of carbon emissions from the perspective of economic development mode. Most existing studies analyze the economic driving force of carbon emissions from the aspects of industrial structure, innovation-driven, and new urbanization(Huang and Yi 2023; Xiao et al. 2023).Lin and Wang (2023), based on empirical analysis, believed that the transfer of high energy consumption and high emission industries was accompanied by a large amount of carbon dioxide transfer, and the industrial transfer of different structures would form different degrees of carbon emission spillover effects. However, Li et al. (2022a) believed that although the path of industrial structure transfer and carbon emissions transfer was similar, the industrial structure transfer and carbon emissions transfer did not produce complete synergy. Zhao et al. (2023) analyzed the data of 252 cities in China and concluded that cooperative green innovation is more effective in reducing carbon emissions than independent green innovation. Xiaomin and Chuanglin (2023) confirmed that the development of urbanization significantly promoted carbon emissions, and the faster the urbanization development, the faster the growth of carbon emissions in provinces. ②It is emphasized to influence carbon emissions through environmental regulation. There are disputes in academic circles about the role of environmental regulation on carbon emission reduction. On the one hand, the "Porter Hypothesis" believes that reasonable environmental regulation can reduce carbon emissions by stimulating technological R&D innovation of enterprises and compensating the cost increase of environmental regulation with innovation(Danish et al. 2020). On the other hand, the "green paradox hypothesis" believes that China's fiscal decentralization governance model will weaken the inhibitory effect of environmental policies on carbon emissions(Zhang et al. 2017). Some scholars also believe that under different degrees of environmental regulation, environmental regulation has different impacts on carbon emission reduction, such as U-shaped (Huang and Tian 2023). ③Dai et al. (2022) test the effect of a series of policies or actions on carbon emission reduction. It can not only reduce the carbon emissions of the region, but also promote the carbon emission reduction effect of the surrounding areas. Meng and Yu (2023) believed that the renewable energy portfolio standard and carbon tax policy promote the carbon emission reduction of China's power industry. Sun and Li (2021) and Zeng et al. (2023) tested the policy effect of low-carbon pilot city construction and high-speed rail construction on carbon emission reduction. (2) Research methods: Most of the existing literatures use spatial econometric model(Li and Li 2020), quantile model(Razzaq et al. 2023), threshold model(Wu et al. 2020), and policy evaluation model(Pan et al. 2020) to study the spatial spillover effect, heterogeneity, interval effect, and policy effect of carbon emissions.

2.2 Studies on the influences brought by UR

The existing research has carried out extensive and in-depth research on the economic, social, and natural environment impacts brought by the application of robots. The development of robots has added new impetus to economic growth. Industrial automation improves the resource allocation efficiency of innovation factors (Chu et al., 2022). Robots may improve productivity growth, but they may have a complex impact on the labor force. Robotic technology has brought positive economic impacts, such as improving productivity, quality, flexibility of multiple industries and breeding new business models, which are expected to have a huge pull on global economic growth(Hubinská 2020). UR could benefit the green economic growth and green technology innovation(Qian et al. 2022; Yin et al. 2022). However, some studies have shown that the development of robot technology will have a negative effect on the labor market. Acemoglu and Restrepo (2020) found that robots may reduce employment and wages by studying the impact of industrial robots on the U.S. labor market (Xie et al., 2022). And the use of industrial robots significantly widened the income gap of urban residents (Zhou et al., 2022). This is mainly reflected in the substitution effect of robots on the labor force in the area. However, Wang et al., (2022) thought robot application has a positive impact on manufacturing employment in the medium and long term.

In addition, some scholars also pay attention to the impact of robot applications on the natural environment, including pollutant emissions, carbon emissions, etc.(Sheng et al., 2022; jiang et al., 2022). Figliozzi and Jennings (2020)found that automatic delivery robots can significantly reduce energy consumption and carbon dioxide emissions in urban areas(Liu et al. 2022). The empirical results show that the use of robot technology significantly reduces the carbon intensity. However, Lange et al., (2020)studied the impact of UR on energy consumption, and found that UR did not save energy, but brought additional energy consumption. In addition, Z. Li & Wang (2022)found that the relationship between digital economy and carbon emissions is inverse-u.

Furthermore, existing studies have also discussed the regional heterogeneity and industry heterogeneity of the effect of UR on carbon emissions. In addition, Z. Li & Wang (2022) believe that there is regional heterogeneity in the impact of the digital economy on carbon emissions. The impact of the digital economy on carbon emissions faces significant resource endowment thresholds, city size thresholds, and innovation capability thresholds. Y. Li et al., (2022) concluded that the emission reduction effect of industrial robots in developed countries is better than that in developing countries. In addition, compared with capital-intensive industries, UR significantly reduces carbon intensity in labor-intensive and technology-intensive industries (Liu et al., 2022). E.-Z. Wang et al. (2022) and Lee et al. (2022) also got similar conclusions. The use of robots has not reached a consensus on the impact of environmental factors such as carbon emissions.

Regarding the specific mechanism of UR affecting carbon emissions, some studies have discussed it from the perspectives of industrial organization, technological innovation, green production efficiency, and energy intensity. Increasing the proportion of non-fossil energy use and optimizing industrial structure are effective mechanisms for digital technology innovation to reduce carbon emission intensity (Wang et al. 2021). Y. Li et al. (2022) found that the application of industrial robots improves productivity, optimizes factor structure, promotes production technology innovation, improves energy efficiency, and reduces carbon intensity. Furthermore, R&D investments curb emission levels and mediate between digitalization and CO2 emissions. Technological innovation has similar direct and moderating effects (Ma et al., 2022). Additionally, UR can promote the marketization of the market and product market. Specifically, the development of technologies such as big data, and the internet has promoted the cross-regional flow of production factors such as talents, capital, and technology, broken down administrative barriers and improved the marketization of production factors. Configurations have a positive impact on carbon reduction. Secondly, robots are widely used in various industries, which significantly promotes the innovation of products and services in various industries, improves the industry's production efficiency, eliminates many backward production capacities, and further upgrades the market structure, thereby reducing carbon emissions. However, the existing research ignores the mediating effect of the marketization.

In general, scholars have conducted a more in-depth discussion on the impact of the application of information technology on carbon emissions, providing theoretical support and experience for subsequent research, but there is still room for further expansion: Firstly, this paper further focuses on the carbon emission reduction effect of UR, and decomposes the carbon emission reduction effect into direct effect part and space spillover effect part with the help of SDM. Secondly, there is little research on the intermediary effect of the degree of marketization on the relationship between the UR and carbon emissions. This paper further establishes an analytical framework through theory and demonstration to explore the carbon emission reduction effect of UR and the intermediary role of marketization in it, which fills the gap of existing research and enriches the existing theoretical mechanisms. This is of great significance to how the government can achieve better carbon emission control through UR in the context of the digital economy. It has important implications for how governments can achieve better carbon emission control through UR in the context of the digital economy.

3.1 Panel regression models

Ehrlich & Holdren (1971) proposed the IPAT model to measure the impact of human activities on the environment. I is the environmental effect related to three main variables: population, affluence, and technology. Dietz & Rosa (1997) created the STIRPAT model and included random effects. The model and its logarithmic form are as follows:

$${\text{I}}_{it}={\text{a}}_{i}{{\text{P}}_{it}}^{b}\text{*}{{\text{A}}_{it}}^{c}\text{*}{{\text{T}}_{it}}^{d}\text{*}{e}_{it}$$

$$\text{I}=\text{P}\text{*}\text{A}\text{*}\text{T}$$

This paper adds the explanatory variables of UR, GDP per capita, degree of marketization, population density, industrial structure, and environmental regulation. The specific formula is as follows, where ${\epsilon }_{i}$is the individual fixed effect, i represents province, and t is the year.

$${C}_{it}={a}_{0}+{a}_{1}{AI}_{it}+{a}_{2}{Z}_{it}+{\tau }_{i}+{u}_{i}+{\epsilon }_{i}$$

Among them, ${C}_{it}$represents the carbon emission of province i in year t. ${AI}_{it}$indicates the number of robots used in province i in year t. ${Market}_{it}$is province's marketization index in year t. ${Z}_{it}$ is a control variable, specifically including PGDP, population density, environmental regulation, and industrial structure. ${u}_{i}$ represents the individual fixed effect, ${\tau }_{i}$ represents the time-fixed effect, and ${\epsilon }_{i}$ represents the error term.

3.2 Panel quantile regressions

The regional differences in the carbon emissions of China's 30 provinces are noticeable and cannot be ignored. Therefore, the carbon emissions of different provinces are heterogeneous, and the traditional panel model may have estimation bias. To accurately identify the mechanism, panel quantile regressions are adopted to measure the different level of impacts of UR and marketization on carbon emissions. Panel quantile regression model with fixed effects is used to obtain more robust and consistent research results.

Quantile regression can depict the impact of dependent variables on different quantile independent variables. It restrains the limitation of least squares regression through using the conditional expectation of the explanatory variable. It gets the regression equation at any quantile of the dependent variable by adding the absolute values of the minimum residual weights. The fixed effect panel quantile regression model estimates parameters by adding penalty items to obtain coefficients. The specific model settings are as follows:

$${Q_{{y_{it}}}}({\tau _k}{x_{it}})=x_{{it}}^{T}\beta ({\tau _{\text{k}}})+{\alpha _i}+{\varepsilon _{it}},i=1, \cdots N,t=1, \cdots ,T$$

${Q_{{y_{it}}}}({\tau _k}{x_{it}})$ indicates the 𝜄k th quantile of UR; x_it is the matrix of Carbon emissions, marketization, PGDP, PD, ER, and IS; 𝛼_i represents the individual fixed effect; 𝜄_k is the quantile; 𝛽(𝜄k) is the 𝜄k th regression coefficients on quantiles.

$$\mathop {\hbox{min} }\limits_{{(\alpha ,\beta )}} \sum\limits_{{k=1}}^{K} {\sum\limits_{{t=1}}^{T} {\sum\limits_{{i=1}}^{N} {{w_k}{\rho _{{\tau _k}}}} } } ({y_{it}} - {\alpha _i} - X_{{it}}^{T}\beta ({\tau _{\text{k}}}))+\lambda \sum\limits_{i}^{N} {\left| {{\alpha _i}} \right|,i} =1, \cdots N,t=1, \cdots ,T$$

Among them, 𝜄 is the quantile, ${\rho _\tau }(u)=u(\tau - I(u<0))$ is the check function, and indicator function is $I(u<0)=\{ \begin{array}{*{20}{c}} 1&{u<0} \\ 0&{u \geqslant 0} \end{array}$, w_k is the weight of the k th quantile and w_k=1/k, 𝜆 (𝜆=1) is the tuning parameter used to penalize the fixed effect 𝛼_i.

3.3 SDM model

Due to the objective economic or policy linkages between regions, the carbon emissions of different regions are interconnected and affect each other, and there is a mutual influence in space. Therefore, it is necessary to consider using spatial econometric models to analyze the impact of various factors on ecological efficiency. SDM is a comprehensive framework and standardized method for spatial economic analysis compared with the spatial lag model (SLM) and spatial error model (SEM) (Li and Wang 2022). This paper uses the following spatial panel Durbin model.

$$Y=\rho WY+\alpha {\iota _N}+X\beta +WX\theta +\varepsilon$$

$\varepsilon \sim N(0,{\delta ^2}{I_n})$

Y is the N × 1 vector of dependent variables; 𝜄N is the vector associated with the constant term 𝛼; X is the N × k matrix of independent variables; 𝛽 is its parameter; 𝜀 represents the error term.

The general form of SDM is:

$${Y}=(I-\rho {W}{)}^{-1}\alpha {\iota }_{n}+(I-\rho {W}{)}^{-1}\left({X}\beta +{W}{X}\theta \right)+(I-\rho {W}{)}^{-1}\epsilon$$

3.4 Mediating effect model

In order to further explore the transmission mechanism of UR's impact on carbon emissions, this paper draws on the research method of Wen and Ye ( 2014). It constructs the following mediation effect model to detect whether UR affects carbon emissions through marketization.

$${C}_{it}={a}_{0}+{a}_{1}{AI}_{it}+{a}_{2}{Z}_{it}+{\tau }_{i}+{u}_{i}+{\epsilon }_{i}$$

$${Market}_{it}={b}_{0}+{b}_{1}{AI}_{it}+{b}_{2}{Z}_{it}+{\tau }_{i}+{u}_{i}+{\epsilon }_{i}$$

$${C}_{it}=c+{c}_{1}{Market}_{it}+{c}_{2}{AI}_{it}+{c}_{3}{Z}_{it}+{\tau }_{i}+{u}_{i}+{\epsilon }_{i}$$

In addition, in this paper, the number of robots used lagged by one period as an instrumental variable to ensure the internal validity of the empirical results.

3.5 Variable and Data

3.5.1 Marketization

The degree of marketization refers to the degree to which the market plays a role in the resource allocation of a country or region. According to (Fan et al., 2003), marketization is China's transition from a planned economy to a market economy, including a series of changes in economics, politics, and culture. For example, the government has reduced its intervention in the economy, the non-state-owned economy has developed, and production factors have accelerated. The marketization index is widely used to describe the relative degree of a region's marketization. Therefore, the marketization index expresses the degree of marketization, and the value range of this variable is 0–10. The larger the number, the higher the degree of marketization.

3.5.2 Variable descriptions

The variables and their specific definitions are shown in Table 1.

Table 1

Variables and Definitions
Variable	Definition	Interpretation
Carbon	Carbon emission	Total carbon emissions
UR	The use of Robots	The number of robots used
Market	Marketization	Market index
PD	Population density	The ratio of the total population to the land area of the province
PGDP	GDP per capita	The ratio of the GDP to the total population of the province
IS	Industry structure	The ratio of the total value of the secondary industry and the tertiary industry to the gross national product
ER	Environmental regulation	Investment in environmental protection as a percentage of GDP

3.5.3 Data Source and Research Sample

This paper further explores the intermediary role of the degree of marketization by studying the UR level and carbon emissions of 30 provinces, autonomous regions, and municipalities in China (excluding Tibet, Hong Kong, Macau, and Taiwan) from 2006 to 2019. The data on the UR level comes from the "White Paper on the Development of Artificial Intelligence." Data related to carbon emissions, industrial structure, per capita GDP, environmental regulation, and population density come from the EPS database. The marketization index is obtained from the "China's Marketization Index Report by Province".

4.1 Panel regression results

Regarding regression coefficient estimation and significance, the regression coefficients of Robots in column 1 and column 4 are significantly negative (Table 2), indicating that UR harms China's total provincial carbon emissions. The continuous improvement of UR can promote the realization of inter-provincial carbon emission reduction. The coefficient of Robots in column 2 is significantly positive, indicating that provinces with a higher level of UR emit less carbon than provinces with a lower level. The sign of the product of the coefficient b1 of Robots in column 2 and the coefficient c1 of Market in column 3 is positive. In contrast, the sign of the coefficient c1 of Robots in column 3 is negative, which proves that the degree of marketization has a masking effect on UR. The amount of masking effect is | b1 c1 /c2 | =0.0035, which shows that the degree of marketization plays a masking effect between UR and carbon emissions, covering the carbon emission reduction effect of UR.

As far as direct effects are concerned, technological breakthroughs in UR will be applied to various industries with the help of information and communication technology infrastructure. After collecting the data of each link, UR can optimize and adjust the workflow of each link from a global perspective, which can significantly reduce waste, improve resource utilization efficiency, and play a positive role in promoting carbon emission reduction. UR's carbon emission reduction effect is mainly reflected in the following aspects: Firstly, in terms of industrial production, the application of UR-related technologies can be used for production scheduling to optimize the efficiency of each process and reduce carbon emissions. Secondly, various production data are collected through the industrial internet of things in intelligent transportation. After processing with the help of relevant algorithms, UR can provide suggestions and even achieve autonomous optimization. Secondly, for example, the application of "green wave speed" and "smart traffic lights" uses artificial intelligence technology to capture road data and analyze real-time road conditions, alleviating road congestion and significantly reducing carbon emissions.

As far as the masking effect is concerned, it also shows that UR can reduce carbon emissions through technological innovation, resource allocation, and structural optimization effects. However, there are still some limitations in China's market-oriented development at this stage. For example, under the assessment pressure of "GDP competition," some provinces are more inclined to protect local high-emission and high-pollution enterprises, which is not unfavorable to the optimal allocation of resources through price signals. In addition, this kind of non-market protection measures is not conducive to creating an innovative environment, is not conducive to the research and development and innovation of green technology, and hurts carbon emission reduction. As a result, the local government's "shelter" for local high-consumption and high-emission enterprises limits UR's carbon emission reduction effect to a certain extent.

Table 2

Results of panel regression model
	Carbon	Market	Carbon	Carbon
UR	-0.139^**	0.0636^**	-0.139^**	-0.156^***
	(0.0604)	(0.0289)	(0.0607)	(0.0563)
PGDP	0.580^*	0.137	0.579^*	0.538^*
	(0.294)	(0.125)	(0.299)	(0.278)
PD	-0.0776	-0.0281	-0.0774	-0.0724
	(0.105)	(0.0462)	(0.107)	(0.0999)
IS	-0.136	0.0599	-0.136	-0.147
	(0.0958)	(0.0651)	(0.0973)	(0.0911)
ER	0.0464^***	-0.0193	0.0466^***	0.0439^***
	(0.0160)	(0.0124)	(0.0164)	(0.0150)
Market			0.00757
			(0.151)
_cons	0.994	0.436	0.991
	(3.226)	(1.204)	(3.231)
N	420	420	420	419
R²	0.767	0.754	0.767	0.768
Standard errors in parentheses
^* p < 0.1, ^ p < 0.05, ^* p < 0.01

4.2 Quantile Regression Analysis

Table 3

− 1 Results of quantile regression model
Quantile	0.1	0.2	0.3	0.4	0.5
UR	− .1248125^***	− .1144286^***	− .1074851^***	− .0986947^***	− .0879758^***
	(.0372691)	(.0299556)	(.0256976)	(.0216495)	(.0197535)
Standard errors in parentheses
^* p < 0.1, ^ p < 0.05, ^* p < 0.01

Table 3

− 2 Results of the quantile regression model
Quantile	0.6	0.7	0.8	0.9
UR	− .079149^***	− .0702867^**	− .0645285^**	− .057325^*
	(.0212462)	(.025107)	(.0284681)	(.0333789)
Standard errors in parentheses
^* p < 0.1, ^ p < 0.05, ^* p < 0.01

As shown in Table 3 − 1 and 3 − 2 above, the coefficients of the panel quantile regression of UR on China's provincial carbon emissions are statistically negative. In addition, as the carbon emission quantile increases, the coefficient size of UR shows a downward trend. This shows that carbon emissions will continue to decline as UR development levels increase; in other words, the development level of UR will have a significant positive effect on China's provincial carbon emissions.

Furthermore, as the quantile of carbon emissions increases, the absolute value of the coefficient of UR gradually decreases, and the negative effect weakens. This shows that in provinces with low carbon emissions, UR's carbon emission reduction effect is more substantial. This is mainly because provinces with low total carbon emissions are mainly concentrated in regions with relatively developed economies or relatively reasonable industrial structures, such as Beijing, Tianjin, and Shanghai. First, a developed economic environment makes it easier to combine UR technology with essential equipment in industrial production and reduce carbon emissions to a greater extent by improving the efficiency of the equipment. Secondly, UR technology has a higher penetration rate in economically developed regions for ordinary citizens' life services, such as internet e-commerce. This has dramatically improved the accuracy and effectiveness of supply and demand docking, optimized resource allocation efficiency, and significantly affected carbon emission reduction. Secondly, market competition in economically developed regions is often more intense, and the market response to the use of artificial intelligence technology will be more evident and rapid, which is conducive to rebuilding a more reasonable market structure, eliminating backward production capacity, and optimizing carbon emission reduction effects.

4.3 Spatial Regression Analysis

Table 4

− 1 Moran index of UR
Year	Proximity distance matrix W1		Geographic distance matrix W2
Year	Statistics	P	Statistics	Statistics
2006	0.198	0.057	0.202	0.013
2007	0.223	0.035	0.224	0.006
2008	0.207	0.049	0.218	0.008
2009	0.183	0.075	0.202	0.013
2010	0.220	0.038	0.219	0.008
2011	0.257	0.018	0.240	0.004
2012	0.331	0.003	0.283	0.001
2013	0.268	0.014	0.252	0.003
2014	0.283	0.010	0.260	0.002
2015	0.283	0.010	0.254	0.003
2016	0.280	0.011	0.249	0.003
2017	0.275	0.012	0.248	0.003
2018	0.268	0.014	0.256	0.002
2019	0.253	0.019	0.245	0.003

Table 4

− 2 Moran index of carbon emissions
Year	Proximity distance matrix W1		Geographic distance matrix W2
Year	Statistics	P	Statistics	P
2006	0.159	0.108	0.143	0.057
2007	0.144	0.132	0.162	0.032
2008	0.238	0.004	0.213	0.043
2009	0.260	0.016	0.315	0.000
2010	0.428	0.000	0.374	0.000
2011	0.309	0.004	0.306	0.000
2012	0.399	0.000	0.315	0.000
2013	0.369	0.001	0.290	0.001
2014	0.371	0.001	0.290	0.001
2015	0.189	0.063	0.235	0.004
2016	0.316	0.004	0.257	0.002
2017	0.262	0.015	0.213	0.009
2018	0.239	0.026	0.237	0.004
2019	0.209	0.048	0.188	0.020

Table 5

LM test results
Test	Proximity distance matrix W1		Geographic distance matrix W2
Test	Statistics	P	Statistics	P
LM Lag	19.7388	0.000	20.4124	0.000
LM Lag (Robust)	3.2145	0.073	10.4653	0.026
LM Error	17.5753	0.000	24.1902	0.000
LM Error (Robust)	10.0510	0.000	5.2431	0.022

Table 4 − 1 and Table 4 − 2 show that Moran's I is positive significantly, except for the proximity distance matrix and geographic matrix used in 2006 and proximity distance matrix in 2006. In addition to the results obtained, other results reject the original assumption that there is no spatial autocorrelation in UR and carbon emissions. There is spatial agglomeration of carbon emissions and UR, that is, the provinces with higher carbon emissions and UR gather in the region and the provinces with lower carbon emissions and UR gather.

Compared with using distance matrix W2, Moran's I value using adjacent matrix W1 is larger, according to the LM Lag and LM Error tests, the SDM model is necessary for analyzing the spatial correlation and spatial spillover effects of UR on carbon emissions. In terms of spatial correlation, in Table 4, the first, second, and third columns show the empirical results of province-fixed effects, time-fixed effects, and two-way fixed effects, respectively; the fourth and fifth columns are mediation effects empirical results. Regarding the spatial correlation of the impact of UR on carbon emissions, the coefficient for robots is negative at the 1% significance level for all SDM models. This shows that UR in a particular province can significantly reduce the carbon emissions of neighboring cities in the province. In addition, the coefficient of PGDP is positive significantly. The economic development in any province is positively correlated with the province's and neighboring provinces' carbon emissions. In addition, population density is negatively correlated with carbon emissions. In addition, the coefficient of industrial structure is negative at the 1% significance level. The coefficient of environmental regulation is positive at the 1% significance level.

As far as the spatial spillover effect of UR on carbon emissions is concerned, based on the results of SDM, the total effect is the sum of direct and indirect effects. The direct effect's regression coefficient is consistent with the explanatory variable of the SDM, and UR can significantly reduce local carbon emissions. In addition, regarding the regression coefficients of indirect effects, UR has significant spatial spillover effects on neighboring regions. Every 1% increase in the level of UR will cause an average decrease of 0.235% in the carbon emissions of adjacent areas. The total effect of UR on the region and its neighbors is 0.356, which indicates that every 1% increase in the level of UR will cause an average decrease of 0.356% in the total carbon emissions of all regions. The level of artificial intelligence has a robust spatial spillover effect.

Specifically, increased levels of UR will positively affect carbon reductions in the region and surrounding areas. It is because UR can enable local enterprises to improve productivity and reduce carbon emissions by optimizing resource allocation and reducing waste, and finally form a "demonstration effect" and a "knowledge spillover effect" brought about by labor flow, driving the surrounding regions. The carbon emission reduction process of provincial enterprises will ultimately affect the total carbon emissions of surrounding provinces. Additionally, the development of local artificial intelligence has expanded the production scale, which can provide clean technology guidance and supervision to enterprises in surrounding provinces and reduce the carbon emissions of affiliated enterprises. Emissions hurt the carbon emissions of surrounding areas. This substantial spatial spillover will significantly improve the ability of regional sustainable and coordinated development.

Per capita GDP can significantly increase the region's carbon emissions, but it has a significant negative spatial spillover effect on the carbon emissions of neighboring provinces. In addition, the industrial structure hurts carbon emissions in the region, but the impact on neighboring provinces is insignificant. The specific performance is that every 1% increase in artificial intelligence will reduce the direct effect and total effect on carbon emissions by 0.124% and 0.248%, respectively. Population density has a significant adverse effect on the province's carbon emissions but has no significant spatial spillover effect on the carbon emissions of neighboring provinces. Specifically, every 1% increase in population density will reduce the direct effect and total effect on carbon emissions by 0.0854% and 0.158%, respectively.

In addition, the mediating effect model shows that the degree of marketization has a significant masking effect on the relationship between UR and carbon emissions. In general, although UR can rely on network effects to allocate elements across space and promote carbon emission reduction by optimizing resource allocation, the degree of marketization development in each province is uneven, and there are still substantial policy barriers to the flow of elements between provinces. However, the carbon emission reduction effect of UR is still limited.

Table 6

results of SDM
	Carbon	Carbon	Carbon	Market	Carbon
Main
UR	-0.0727^***	-0.470^***	-0.113^***	0.0466^***	-0.118^***
	(0.0212)	(0.0259)	(0.0207)	(0.0129)	(0.0211)
PGDP	0.712^***	-0.0256	0.680^***	0.151^**	0.669^***
	(0.105)	(0.0756)	(0.102)	(0.0640)	(0.103)
PD	-0.0746^***	-0.256^***	-0.0849^***	-0.0229	-0.0832^***
	(0.0281)	(0.0533)	(0.0271)	(0.0171)	(0.0272)
IS	-0.119^***	-0.592^***	-0.121^***	0.0448	-0.126^***
	(0.0453)	(0.0701)	(0.0462)	(0.0291)	(0.0464)
ER	0.0426^***	0.236^***	0.0438^***	-0.0161^**	0.0455^***
	(0.0103)	(0.0347)	(0.0103)	(0.00650)	(0.0104)
Market					0.0788
					(0.0790)
Wx
UR	-0.0181	-0.0139	-0.187^***	0.0857^***	-0.198^***
	(0.0283)	(0.0538)	(0.0399)	(0.0252)	(0.0422)
PGDP	-0.0145	-0.138	-0.545^***	0.144	-0.566^***
	(0.136)	(0.137)	(0.180)	(0.106)	(0.181)
PD	0.00383	-0.641^***	-0.0409	0.0831^**	-0.0427
	(0.0595)	(0.131)	(0.0593)	(0.0374)	(0.0612)
IS	0.0190	-1.053^***	-0.0862	0.269^***	-0.113
	(0.0541)	(0.153)	(0.0985)	(0.0623)	(0.101)
ER	-0.0403^**	0.141^**	-0.00842	-0.0141	-0.00558
	(0.0165)	(0.0662)	(0.0212)	(0.0134)	(0.0218)
Market					0.0221
					(0.155)
ρ	0.352^***	0.354^***	0.356^***	0.374^***	0.358^***
N	420	420	420	420	420
R²	0.340	0.360	0.331	0.246	0.330
Standard errors in parentheses
^* p < 0.1, ^ p < 0.05, ^* p < 0.01

Table 7

Direct and indirect impacts of UR in SDM
	Carbon	Carbon	Carbon	Market	Carbon
Direct
UR1	-0.0769^***	-0.471^***	-0.121^***	0.0577^***	-0.126^***
	(0.0204)	(0.0261)	(0.0203)	(0.0130)	(0.0206)
UR2	-0.106^***	0.492^***	-0.125^***	0.0543^***	-0.127^***
	(0.0210)	(0.0259)	(0.0209)	(0.0135)	(0.0215)
Indirect
UR1	-0.0623^*	-0.0528	-0.235^***	0.157^***	-0.249^***
	(0.0366)	(0.0418)	(0.0477)	(0.0378)	(0.0503)
UR2	-0.0249	-0.0246	-0.242^***	0.164^***	-0.243^***
	(0.0445)	(0.0704)	(0.0631)	(0.0454)	(0.0635)
Total
UR1	-0.139^***	0.418^***	-0.356^***	0.215^***	-0.375^***
	(0.0375)	(0.0429)	(0.0521)	(0.0425)	(0.0504)
UR2	-0.131^***	0.467^***	-0.366^***	0.218^***	-0.370^***
	(0.0428)	(0.0763)	(0.0657)	(0.0473)	(0.0618)
Standard errors in parentheses
^* p < 0.1, ^ p < 0.05, ^* p < 0.01

This paper takes 30 provinces in China as research samples and uses the fixed effect model and SDM model to explore the impact of UR on carbon emissions and its spatial spillover effects; further, the degree of marketization is introduced into the mediating effect model to verify the degree of marketization. The specific mechanism of action on the relationship between UR and carbon emissions. Conclusions are as follows.

Firstly, UR can significantly reduce carbon emissions. Moreover, in provinces with low carbon emissions, artificial intelligence's carbon emission reduction effect is more substantial. First, UR can empower production facilities through information technology, improve process efficiency, reduce resource waste, and achieve carbon emission reduction. Secondly, with the help of communication technologies such as the Internet, UR can achieve more accurate supply and demand docking in product and factor markets, optimize resource allocation, and achieve positive carbon emission reduction effects. Finally, UR will reshape the market competition pattern through the scale effect, eliminate a batch of backward production capacity, and reduce carbon emissions.

Secondly, UR will not only have a positive effect on carbon emission reduction in the region but also have a significant spatial spillover effect on carbon emission reduction in surrounding provinces. This may be due to the "demonstration effect" and the "knowledge spillover effect" brought about by labor mobility. Artificial intelligence empowers local enterprises to improve production efficiency and reduce carbon emissions and finally forms a demonstration effect, which drives surrounding enterprises to reduce carbon emissions actively. In addition, the flow of labor between neighboring provinces will bring clean technology spillovers, and the carbon emissions of neighboring provinces will decrease.

Thirdly, UR has a positive carbon emission reduction effect, but the degree of marketization has a masking effect. Although UR has promoted the flow of factor markets and product markets and increased the degree of marketization, the degree of marketization has increased carbon emissions. This may be because some provinces in China are more inclined to provide political protection for local polluting enterprises due to the more significant pressure of GDP assessment, which interferes with the regular market order and increases carbon emissions.

Based on the above, this paper puts forward the following policy recommendations:

Firstly, in the context of an increasingly aging population and a low birth rate, formulate relevant policies to promote the high-quality development of the robot industry and use the "robot dividend" to create new economic development points. The upgrading of the robot industry promotes production efficiency and improves social welfare. For example, in the elderly care industry, long-term care, etc., the intervention of artificial intelligence can not only alleviate the labor shortage in related industries but also achieve higher service quality and gradually release the "robot dividend."

Secondly, actively promote the development of artificial intelligence-enabled technologies related to carbon emission reduction. UR is used to reduce carbon emissions and has the function of predicting and monitoring emissions. Artificial intelligence-related technologies can be used to predict future emissions based on existing carbon emission data; in addition, UR can track carbon footprint data in real-time, realizing carbon emission monitoring and monitoring of the entire process from production and sales to operation and maintenance.

At a deeper level, better use of UR to promote the coordinated development of regional carbon emission reduction. UR has a robust spatial spillover effect on carbon emissions in surrounding provinces. Therefore, the government can formulate corresponding talent and technology exchange policies among surrounding provinces to facilitate the cross-regional flow of production factors such as clean technology and equipment, thereby promoting carbon reduction—regional coordinated development of the platoon.

Data availability

All datasets, models, or codes used in this study are available from the corresponding author by resonable request.

Ethical approval

The data in this paper are from public databases, and the study does not involve humans or animals, so the paper does not require ethical approval.

Consent to participate

All authors of the article consent to participate.

Consent for publication

All authors of the article consent to publish.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Competing Interests

The authors have no relevant financial or non-financial interests to disclose.

Author Contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Yang Zhou, Yang Nie, Liangen Zeng and Wenchu Bao. The first draft of the manuscript was written by Hankun Wang, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Dai S, Qian Y, He W, et al (2022) The spatial spillover effect of China’s carbon emissions trading policy on industrial carbon intensity: Evidence from a spatial difference-in-difference method. Structural Change and Economic Dynamics 63:139–149. https://doi.org/10.1016/j.strueco.2022.09.010
Danish, Ulucak R, Khan SU-D, et al (2020) Mitigation pathways toward sustainable development: Is there any trade-off between environmental regulation and carbon emissions reduction? Sustainable Development 28:813–822. https://doi.org/10.1002/sd.2032
Dietz T, Rosa EA (1997) Effects of population and affluence on CO ₂ emissions. Proc Natl Acad Sci USA 94:175–179. https://doi.org/10.1073/pnas.94.1.175
Ehrlich PR, Holdren JP (1971) Impact of Population Growth. Science, New Series 171:1212–1217
Figliozzi M, Jennings D (2020) Autonomous delivery robots and their potential impacts on urban freight energy consumption and emissions. Transportation Research Procedia 46:21–28. https://doi.org/10.1016/j.trpro.2020.03.159
Fan Gang, Wang Xiaolu, Zhang Liwen, Zhu Hengpeng (2003) The relative progress of marketization in China by region. Economic Studies 9–18 + 89 (In Chinese)
Huang H, Yi M (2023) Impacts and mechanisms of heterogeneous environmental regulations on carbon emissions: An empirical research based on DID method. Environmental Impact Assessment Review 99:107039. https://doi.org/10.1016/j.eiar.2023.107039
Huang X, Tian P (2023) How does heterogeneous environmental regulation affect net carbon emissions: Spatial and threshold analysis for China. Journal of Environmental Management 330:117161. https://doi.org/10.1016/j.jenvman.2022.117161
Hubinská Z (2020) PRACTICAL USE OF INNOVATIVE DIDACTIC TOOLS IN DANCE EDUCATION
Lange S, Pohl J, Santarius T (2020) Digitalization and energy consumption. Does ICT reduce energy demand? Ecological Economics 176:106760. https://doi.org/10.1016/j.ecolecon.2020.106760
Lee C-C, Qin S, Li Y (2022) Does industrial robot application promote green technology innovation in the manufacturing industry? Technological Forecasting and Social Change 183:121893. https://doi.org/10.1016/j.techfore.2022.121893
Li J, Li S (2020) Energy investment, economic growth and carbon emissions in China—Empirical analysis based on spatial Durbin model. Energy Policy 140:111425. https://doi.org/10.1016/j.enpol.2020.111425
Li M, Li Q, Wang Y, Chen W (2022a) Spatial path and determinants of carbon transfer in the process of inter provincial industrial transfer in China. Environmental Impact Assessment Review 95:106810. https://doi.org/10.1016/j.eiar.2022.106810
Li Y, Zhang Y, Pan A, et al (2022b) Carbon emission reduction effects of industrial robot applications: Heterogeneity characteristics and influencing mechanisms. Technology in Society 70:102034. https://doi.org/10.1016/j.techsoc.2022.102034
Li Z, Wang J (2022) The Dynamic Impact of Digital Economy on Carbon Emission Reduction: Evidence City-level Empirical Data in China. Journal of Cleaner Production 351:131570. https://doi.org/10.1016/j.jclepro.2022.131570
Lin B, Wang C (2023) Does industrial relocation affect regional carbon intensity? Evidence from China’s secondary industry. Energy Policy 173:113339. https://doi.org/10.1016/j.enpol.2022.113339
Liu J, Liu L, Qian Y, Song S (2022) The effect of artificial intelligence on carbon intensity: Evidence from China’s industrial sector. Socio-Economic Planning Sciences 83:101002. https://doi.org/10.1016/j.seps.2020.101002
Ma Q, Tariq M, Mahmood H, Khan Z (2022) The nexus between digital economy and carbon dioxide emissions in China: The moderating role of investments in research and development. Technology in Society 68:101910. https://doi.org/10.1016/j.techsoc.2022.101910
Meng X, Yu Y (2023) Can renewable energy portfolio standards and carbon tax policies promote carbon emission reduction in China’s power industry? Energy Policy 174:113461. https://doi.org/10.1016/j.enpol.2023.113461
Pan X, Li M, Wang M, et al (2020) The effects of a Smart Logistics policy on carbon emissions in China: A difference-in-differences analysis. Transportation Research Part E: Logistics and Transportation Review 137:101939. https://doi.org/10.1016/j.tre.2020.101939
Qian Y, Liu J, Shi L, et al (2022) Can artificial intelligence improve green economic growth? Evidence from China. Environ Sci Pollut Res 30:16418–16437. https://doi.org/10.1007/s11356-022-23320-1
Razzaq A, Sharif A, Afshan S, Li CJ (2023) Do climate technologies and recycling asymmetrically mitigate consumption-based carbon emissions in the United States? New insights from Quantile ARDL. Technological Forecasting and Social Change 186:
Sun L, Li W (2021) Has the opening of high-speed rail reduced urban carbon emissions? Empirical analysis based on panel data of cities in China. Journal of Cleaner Production 321:128958. https://doi.org/10.1016/j.jclepro.2021.128958
Wang E-Z, Lee C-C, Li Y (2022) Assessing the impact of industrial robots on manufacturing energy intensity in 38 countries. Energy Economics 105:105748. https://doi.org/10.1016/j.eneco.2021.105748
Wang L, Chen Y, Ramsey TS, Hewings GJD (2021) Will researching digital technology really empower green development? Technology in Society 66:101638. https://doi.org/10.1016/j.techsoc.2021.101638
Wu H, Xu L, Ren S, et al (2020) How do energy consumption and environmental regulation affect carbon emissions in China? New evidence from a dynamic threshold panel model. Resources Policy 67:101678. https://doi.org/10.1016/j.resourpol.2020.101678
Wen Zhonglin, Ye Baojuan (2014) Mediation effect analysis: Methodology and model development. Advances in Psychological Science 22: 731–745. (In Chinese)
Xiao Y, Huang H, Qian X-M, et al (2023) Can new-type urbanization reduce urban building carbon emissions? New evidence from China. Sustainable Cities and Society 90:104410. https://doi.org/10.1016/j.scs.2023.104410
Xiaomin G, Chuanglin F (2023) How does urbanization affect energy carbon emissions under the background of carbon neutrality? Journal of Environmental Management 327:116878. https://doi.org/10.1016/j.jenvman.2022.116878
Yin K, Cai F, Huang C (2022) How does artificial intelligence development affect green technology innovation in China? Evidence from dynamic panel data analysis. Environ Sci Pollut Res 30:28066–28090. https://doi.org/10.1007/s11356-022-24088-0
Zeng S, Jin G, Tan K, Liu X (2023) Can low-carbon city construction reduce carbon intensity༟Empirical evidence from low-carbon city pilot policy in China. Journal of Environmental Management 332:117363. https://doi.org/10.1016/j.jenvman.2023.117363
Zhang K, Zhang Z-Y, Liang Q-M (2017) An empirical analysis of the green paradox in China: From the perspective of fiscal decentralization. Energy Policy 103:203–211. https://doi.org/10.1016/j.enpol.2017.01.023
Zhao Y, Zhao Z, Qian Z, et al (2023) Is cooperative green innovation better for carbon reduction? Evidence from China. Journal of Cleaner Production 394:136400. https://doi.org/10.1016/j.jclepro.2023.136400

Download PDF

Journal Publication

published 18 Oct, 2023

Read the published version in Environmental Science and Pollution Research →

Editorial decision: Major Revision
12 Jun, 2023
Reviewers agreed at journal
10 May, 2023
Reviewers invited by journal
10 May, 2023
Editor assigned by journal
04 Apr, 2023
First submitted to journal
28 Mar, 2023

You are reading this latest preprint version

How does the use of robots promote carbon reduction?——Spatial correlation and heterogeneity analysis

Status:

Journal Publication

Version 1

Abstract

1 Introduction

2 Literature Review

2.1 Studies on driving factors of carbon emission reduction

2.2 Studies on the influences brought by UR

3 Methodology

3.1 Panel regression models

3.2 Panel quantile regressions

3.3 SDM model

3.4 Mediating effect model

3.5 Variable and Data

3.5.1 Marketization

3.5.2 Variable descriptions

3.5.3 Data Source and Research Sample

4 Results and Discussion

4.1 Panel regression results

4.2 Quantile Regression Analysis

4.3 Spatial Regression Analysis

5 Conclusion and Policy Recommendations

Declarations

References

Status:

Journal Publication

Version 1