The Impact of Information Infrastructure Construction on Carbon Emissions

Abstract

As the economy continues to grow, greenhouse gas emissions are increasing, and people are paying more attention to issues related to carbon emissions. The construction of information infrastructure has also become an important support for development in the new era. Therefore, to explore whether information infrastructure construction can reduce carbon emissions, this paper conducts a series of tests based on panel data from 30 provinces in China from 2010 to 2019. The empirical results show that, in addition to directly reducing carbon emissions, the construction of information infrastructure can also affect carbon emissions through technological innovation. According to the heterogeneity test, we find that the construction of information infrastructure has better emission reduction effects for the eastern region, provinces with a higher technological innovation level, and provinces with a higher carbon emission intensity. The results of this paper provide strong support for the mitigation of global warming.

1. Introduction

With the development of China’s industrial economy, the massive consumption of fossil fuels has increased emissions of carbon dioxide, one of the greenhouse gases, making China the world’s largest carbon emitter. According to the World Energy Yearbook 2021, China’s carbon emissions rose from 8.83 billion tonnes to 9.90 billion tonnes between 2011 and 2020. In 2021, China’s carbon dioxide emissions reached 10,523 million tons, accounting for 45% of global carbon dioxide emissions. At the same time, global warming caused by massive carbon emissions is a hot environmental issue of concern to all countries, causing glacial melting, extreme weather, and even local species extinction, posing a serious challenge to human society. The State of the Global Climate 2022 report released by the World Meteorological Organization shows that the global average temperature over the past eight years is the highest ever recorded. Compared to pre-industrial temperatures, the global average temperature in 2022 has increased by 1.15 °C. The change in carbon dioxide emissions per capita for China as a whole and by region from 2010 to 2019 is illustrated in Figure 1. In response to this, reducing carbon dioxide emissions and improving the efficiency of carbon emissions has become a major challenge for many countries in order to combat global warming. In accordance with the Paris Agreement, the Chinese government has set a 2030 deadline for reaching a peak in carbon dioxide emissions and a 60–65% reduction in carbon emissions per unit of GDP. Implementing carbon emission reductions and improving carbon emission efficiency has become an important strategy for ecological civilization construction and green low-carbon development.

Under the new development pattern proposed by China of “a large domestic cycle as the main body and a dual domestic and international cycle to promote each other”, information infrastructure construction has become an important carrier of the digital economy. Moreover, technologies such as cloud computing, the Internet of Things, and machine learning are developing at the same time [1], and information infrastructure construction has the potential to be a powerful weapon to improve carbon emissions. Therefore, we are increasingly concerned about the impact of information infrastructure construction on carbon emissions. It can reduce carbon dioxide emissions by improving energy efficiency, upgrading and transforming the industrial structure of enterprises, and improving technological innovation capabilities. It is of great theoretical and practical significance to explore the impact of information infrastructure construction on the reduction of carbon emissions.

Researchers have studied carbon emissions mainly from the following aspects. The first category examines how carbon emissions are measured and analyzes the factors that influence them. A great deal of research has been done on methods for measuring carbon emissions in different regions or sectors [2,3,4,5]. Economic growth, technological progress, and demographics all have a significant impact on carbon emissions [6,7,8]. At the same time, government regulations and environmental policies can be effective in curbing carbon emissions [9,10,11]. Changes in urban construction also have an important impact on carbon emissions [12,13]. The second category is the impact of Internet development on carbon emissions in the development of information infrastructure. With the development of the Internet, industrial convergence disseminates knowledge and information to traditional industries through the penetration of internet digital technology, reducing carbon emissions. In addition, it can also increase the incentive for enterprises to invest in information and curb high-pollution operations [14,15]. The development of the internet facilitates the development of carbon trading markets, thus achieving a reduction in carbon emissions [16]. In the carbon trading market, internet technology solves the technical problems associated with emissions detection, reporting, and verification. The third category is the impact of information infrastructure construction on carbon emissions in the context of the digital economy. The development of the digital economy can promote the digitization and structural upgrading of industries, effectively improving the traditional mode of production organization [17,18,19]. As a result, emerging industries will make more use of non-polluting clean production factors for production and reduce carbon dioxide emissions. The newly created products or services drive the transformation of the consumption structure [20]. Finally, the development of the digital economy can also promote changes in the energy mix through technological innovation, thereby reducing carbon dioxide emissions [21].

The research implications of this paper are as follows. Firstly, we measure the development level index of information infrastructure construction from two major aspects and explore its relationship with carbon emissions accordingly. In addition, we discuss the mechanism of the impact of information infrastructure construction on carbon emissions and use the intermediary effect to analyze the mechanism from the perspective of technological innovation, providing a new way of thinking about carbon emission reduction. Considering that different regions, different carbon emission intensities, and different degrees of technological innovation in China have different impacts on carbon emission intensity, we conducted a heterogeneity analysis to broaden the scope of the study and enrich the research direction, providing some theoretical support for the impact of information infrastructure construction on carbon emission in China. Moreover, we can provide some information for policymakers. It is of practical meaning to determine the level of development of information infrastructure construction, carbon emissions, and technological progress in China; the role of information infrastructure construction in promoting technological innovation; and the differential effects of information infrastructure construction on carbon emission reductions in different regions, different carbon emission intensities, and different levels of technological innovation in China so as to propose more targeted improvement measures to improve the carbon emission reduction effect in China.

The rest of the article is structured as follows. In Section 2, we compare the direct and indirect impacts of information infrastructure construction on carbon emissions. Empirical data and econometric modeling methods are introduced in Section 3. Section 4 introduces the empirical results and analyzes them accordingly, including fundamental regression, mediation effect testing, and robustness testing. Section 5 presents the results of a heterogeneity analysis under the three classification scenarios. Section 6 presents the key conclusions and recommendations.

2. Literature Review

2.1. Direct and Indirect Impact of Information Infrastructure Development on Carbon Emissions

The construction of information infrastructure, represented by 5G, artificial intelligence, and big data, embodies eco-friendly attributes in an all-round way and is developing toward the greening, colocalization, and modernization of human society. Therefore, the role of information infrastructure construction in reducing carbon dioxide emissions should not be ignored. In 2002, China put forward the strategy of “integrated development of information and industrialization”, which aims to promote the green transformation of traditional industries through information and communication technologies (ICT). Zhou et al. (2016) and Li et al. (2019) found that by promoting the development of emerging industries and upgrading industrial structures, information infrastructure construction will further promote the agglomeration of emerging industries based on the Internet, thus reducing the proportion of traditional high-polluting industries and reducing carbon dioxide emissions [22,23]. Yang et al. (2022) used panel data from 289 cities in China from 2011 to 2017 to explore the impact and mechanism of information infrastructure on urban carbon emission intensity. The study found that information infrastructure significantly reduced the carbon emission intensity of cities [24]. Zhang et al. (2022) discussed and analyzed the impact and mechanism of information on the infrastructure on air pollution based on the data from 31 provinces in China from 2013 to 2020. The results showed that information infrastructure could effectively improve air quality, that is, reduce carbon dioxide emissions [25]. Kou et al. (2022) used the Malmquist–Luenberger index model to measure the carbon emission efficiency of Chinese cities from 2004 to 2017 and used a panel quantitative regression model to analyze the impact of Internet infrastructure on carbon emission efficiency. The results show that the Internet infrastructure has a significant inhibitory effect on the carbon emission intensity under the market segment [26]. Dong et al. (2022) used panel data from 281 prefecture-level cities in China from 2003 to 2018 to consider the broadband China policy as a quasi-natural experiment of information infrastructure and conducted a difference-in-difference (DID) analysis. The results showed that information infrastructure significantly improved urban greenhouse gas emission performance [27].

However, some researchers believe that the construction of information infrastructure will lead to an increase in carbon emissions. Salahuddin and Alam (2015) examined the empirical relationship between internet use, electricity consumption, and economic growth using annual time series data for the period 1985–2012 for Australia. In the long term, both internet use and economic growth stimulated electricity consumption in Australia, and this positive relationship suggests that Australia has yet to realize energy efficiency gains from the expansion of ICTs [28]. Hamdi et al. (2015) explored the relationship between electricity consumption, foreign direct investment, capital, and economic growth. They find that the construction of infrastructure leads to an increase in electricity consumption, which in turn leads to an increase in carbon emissions [29].

2.2. The Impact of Information Infrastructure Construction on Carbon Emissions through Technological Innovation

There are many factors that influence carbon emissions, but with the advancement of science and technology, more and more domestic and foreign researchers are focusing on the impact of technological innovation on reducing carbon emissions. Porter’s hypothesis (1996) holds that technological innovation can improve environmental quality by improving enterprise productivity and offsetting environmental protection costs [30]. Fan et al. (2006) used the STIRPA model to study the impact of population, affluence, and technology on total carbon dioxide emissions in countries with different income levels over the period 1975–2000. The results showed that the effect of technology on total carbon dioxide emissions varied across different levels of development [31]. Apergis et al. (2013) investigated the manufacturing sectors of Germany, France, and the United Kingdom during 1998–2011 and found that R&D expenditure had a significant effect on reducing carbon dioxide emissions [32]. Font et al. (2014) studied the contribution of technological innovation to environmental pressure using a decomposition analysis based on LCA and found that technological innovation could reduce the emissions of carbon dioxide and nitrogen oxides [33]. Acemoglu et al. (2014) introduced technological change into the general equilibrium model of climate change and found that technological change could affect carbon emissions [34]. Shaikh et al. (2018) used the autoregressive distributed lag (ARDL) model to study the dynamic impacts of technological innovation, financial development, economic growth, and energy on carbon dioxide emissions in China during 1980–2017. It turns out that technological innovations can reduce carbon dioxide emissions [35]. Anwar et al. (2021) simulated the macroeconomic determinants of carbon dioxide emissions in G7 countries from 1996 to 2018 and found that technological innovation could hinder the level of carbon dioxide emissions [36]. Liu et al. (2022) studied the mechanism of action and spillover effects of technological innovation on carbon emission reduction using panel data for 30 Chinese provinces from 2008 to 2019. The researchers believed that technological innovation has a direct negative impact on carbon emissions at the provincial level and also has a spatial spillover effect on neighboring provinces [37].

3. Materials and Methods

3.1. Variables

Based on the availability of data, the sample selected for this paper was a panel of 30 Chinese provinces from 2010 to 2019, excluding Tibet, Hong Kong, Macao, and Taiwan. Among them, the carbon emission intensity data comes from the China Carbon Accounting Database and the China Statistical Yearbook, and the information infrastructure construction index data comes from the China Statistical Yearbook. Data on technological innovation of intermediate variables are from the China Statistical Yearbook, and control variables are from the China Energy Statistical Yearbook, the China Statistical Yearbook, the China Environment Statistics Yearbook, and the Provincial Statistical Yearbook [38,39,40,41].

3.1.1. Explained Variable

The method of calculating the carbon emission intensity (CEI) is based on the existing widely used calculation method with the formula:

$$\begin{array}{c}\hfill CEI={\displaystyle \frac{CDE}{GDP}}\end{array}$$

where CDE represents regional carbon dioxide emissions, and GDP represents regional gross domestic product.

3.1.2. Core Explanatory Variable

Information infrastructure construction (IIC) refers to the communication network foundation represented by 5G, Internet of Things, industrial Internet, and satellite Internet, and the infrastructure of new technologies represented by artificial intelligence, cloud computing, and blockchain. When constructing the information infrastructure construction index, whether it is represented by a single variable or constructed by the number of enterprises in some related fields, the results cannot accurately cover the foundation and application levels of information infrastructure construction. Therefore, the following four indicators are used, taking into account the development process of China’s information infrastructure construction and the existing construction methods (

Table 1. Indicators used to build the information infrastructure construction.

Variables	Unit	Data Source
The length of long-distance fiber-optic cable lines	Million kilometres	National Bureau of Statistics of China
The Internet penetration rate	The proportion of Internet users to the population at the end of the year	China Internet Development Status Report and Internet Development Report by province
The mobile phone penetration rate	The number of mobile phone users per 100 people	National Bureau of Statistics of China
The proportion of Internet broadband access ports	The number of Internet broadband access ports per capita	National Bureau of Statistics of China

). The length of long-distance fiber-optic cable lines and the proportion of Internet broadband access ports (the number of Internet broadband access ports per capita) represent the infrastructure; the mobile phone penetration rate (the number of mobile phone users per 100 people) and the Internet penetration rate (the proportion of Internet users to the population at the end of the year) represent the application level. The above four indicators are used to construct the Information Infrastructure Construction Index using the entropy weighting method (IIC) [38,42].

3.1.3. Mediating Variable

Technological innovation (lnTI) is one of the most direct and effective ways to reduce carbon emissions, as it will directly change the way people produce and live, resulting in some reduction in carbon emissions. This paper uses the logarithm of the number of domestic patent applications by province as a proxy for technological innovation.

3.1.4. Control Variables

The control variables are as follows:

(1) Foreign direct investment (lnFDI): Among the effects of this variable on carbon emissions, some scholars believe that FDI in factory construction will aggravate domestic carbon dioxide emissions, while some others believe that FDI can introduce some advanced technologies or promote industrial structure upgrading, thus reducing domestic carbon dioxide emissions. It is expressed as the logarithm of imports and exports of foreign-invested enterprises in each province.

(2) Energy consumption (lnEnergy): The direct source of carbon dioxide is the consumption of energy such as coal, natural gas, LPG, etc. Therefore, the more energy is consumed, the more carbon dioxide is produced. It is expressed as the logarithm of the total energy consumption in each province.

(3) Openness (lnOpen): The degree of openness has been shown in many studies to have a significant impact on carbon dioxide emissions. On the one hand, openness can promote economic development and thus technological innovation and reduce carbon emissions. On the other hand, for some developing countries with low environmental standards and inadequate environmental regulations, the higher the degree of openness, the higher carbon dioxide emissions. It is expressed as the logarithm of total trade imports and exports divided by the regional GDP.

(4) The level of economic development (lnGDP): This has a twofold effect on carbon emissions. On the one hand, economic development promotes the intensity of industries, economic activities, and people’s lives, which in turn increases carbon emissions. On the other hand, when economic development is followed by economic development, it can, to a certain extent, promote the government to strengthen environmental management, technological innovation, etc., thus reducing carbon emissions. It is expressed as the logarithm of regional GDP.

(5) Environmental regulation (ER): Some scholars argue that the monitoring of government environmental regimes can reduce carbon emissions. Some scholars also believe that environmental regulations are used for technological innovation to produce a backlash effect, which, in turn, increases carbon emissions. It is expressed in terms of industrial sulfur dioxide emissions, industrial wastewater emissions, and industrial dust emissions using the entropy method of calculation.

Table 2. Variables and descriptive statistics.

Variables	Symbol	Mean	Min.	Max.	Standard Deviation	Data Source
Carbon emission intensity	$CEI$	2.397	0.199	12.832	2.282	China Carbon Data Accounting Database
Information infrastructure construction	$IIC$	−1.914	−2.957	−1.106	0.402	Same as Section 3.1.2
Technological innovation	$lnTI$	10.532	6.400	13.602	1.453	National Bureau of Statistics of China
Foreign direct investment	$lnFDI$	9.411	7.214	10.631	0.657	National Bureau of Statistics of China
Energy consumption	$lnEnergy$	9.680	7.042	11.573	0.894	China Energy Statistics Yearbook
Openness	$lnOpen$	16.055	8.330	20.199	2.515	National Bureau of Statistics of China and provincial statistical offices
Level of economic development	$lnGDP$	−3.272	−9.041	−0.035	1.927	National Bureau of Statistics of China and provincial statistical offices
Environmental Regulation	$ER$	0.524	0	2.585	00.535	China Statistical Yearbook and China Environmental Statistics Yearbook

gives the description of the main variables in this article.

Prior to the empirical tests, we first conducted correlation tests on the variables to determine whether there was multiple collinearity between the variables. The results of the test are shown in

Table 3. Correlation test results.

	IIC	CEI	lnFDI	lnEnergy	lnOpen	lnTI	lnGDP	ER
$IIC$	1
$CEI$	−0.305	1
$lnFDI$	0.264	−0.447	1
$lnEnergy$	0.122	−0.057	0.315	1
$lnOpen$	0.206	−0.374	0.772	0.143	1
$lnTI$	0.361	−0.516	0.619	0.401	0.423	1
$lnGDP$	0.303	−0.455	0.573	0.576	0.347	0.749	1
$ER$	−0.005	0.101	0.162	0.685	0.02	0.204	0.368	1

. The results of the test show that the correlation coefficient between the variables did not exceed 0.8, i.e., there was no multicollinearity between the variables.

3.2. Model Construction

3.2.1. Basic Regression Model

To test the impact of information infrastructure on carbon emission intensity, we chose to use a fixed effects model. Additionally, to control for regional macroeconomic differences and areas that do not vary over time, we used a double fixed effects model; i.e., we fixed both the year and the province. The basic regression model is set up as follows.

$$\begin{array}{c}\hfill Y_{it}=\mu +\beta X_{it}+{\lambda }_1{Z}_{1it}+{\lambda }_2{Z}_{2it}+{\lambda }_3{Z}_{3it}+{\lambda }_4{Z}_{4it}+{\lambda }_5{Z}_{5it}+{\lambda }_6{Z}_{6it}{\alpha }_i+{\gamma }_t+{ϵ}_{it}\end{array}$$

(1)

where $Y_{it}$ represents the carbon emission intensity of province i in year t, $X_{it}$ represents the information infrastructure construction of province i in year t, $Z_{1it},\dots ,Z_{6it}$ denotes a set of control variables, and $\alpha$ denotes the intercept term. $\mu$ represents the coefficient of the control variable, and ${\lambda }_1,\dots ,{\lambda }_6$ represents the coefficient of each control variable. ${\alpha }_i$ and ${\gamma }_t$ represent the fixed effect of the province and the fixed effect of time. ${ϵ}_{it}$ is the error term. The coefficient $\beta$ is an indication of the magnitude of the impact of information infrastructure on carbon emissions intensity. A significant negative $\beta$ indicates a reduction in carbon emission intensity due to enhanced information infrastructure, i.e., an effective reduction in carbon emission production per unit of GDP. Otherwise, if $\beta$ is not significant or the coefficient is opposite to the expected result, it indicates that the influence of information infrastructure construction on carbon emission intensity is not very ideal compared with the expected result.

3.2.2. Intermediary Effect Model

The basic logic of the intermediate variable is as follows. When we analyze the effect of the explanatory variable X on the explained variable Y, if the explanatory variable X affects the explained variable Y through the variable M, then the variable M is the intermediate variable. Thus, when there is an intermediary effect, the effect of the explanatory variable on the explained variable ($c^{\prime }$) should be smaller than in a model without the intermediate variable (c). The path of the effect of the intermediate variable is shown in Figure 2.

In Figure 2, coefficient c represents the total effect of the explanatory variable X on the explained variable Y. Coefficient a represents the effect of the explanatory variable X on the intermediary variable M. Coefficient b represents the effect of the intermediary variable M on the explained variable Y after controlling the influence of the explanatory variable X. Coefficient $c^{\prime }$ represents the direct effect of the explanatory variable X on the explained variable Y after controlling the influence of the intermediary variable M, and e is the error term.

The analysis of the theoretical influence between information infrastructure construction, carbon emission intensity, and technological innovation shows that technological innovation has a positive effect on the reduction of carbon emissions. Therefore, we want to deeply understand the relationship between the three and further test whether technological innovation exists as an intermediary effect in the process of the influence of information infrastructure construction on carbon emission intensity. Based on the research needs, we choose the mediation effect research method to study the mediation effect of technological innovation. The model contains explanatory variable X, explained variable Y, and intermediary variable M.

$$\begin{array}{c}\hfill Y_{it}=c{X}_{it}+{\alpha }_1{Z}_{1it}+{\alpha }_2{Z}_{2it}+{\alpha }_3{Z}_{3it}+{\alpha }_4{Z}_{4it}+e_1\end{array}$$

(2)

$$\begin{array}{c}\hfill M_{it}=a{X}_{it}+{\beta }_1{Z}_{1it}+{\beta }_2{Z}_{2it}+{\beta }_3{Z}_{3it}+{\beta }_4{Z}_{4it}+e_2\end{array}$$

(3)

$$\begin{array}{c}\hfill Y_{it}=c^{\prime }{X}_{it}+b{M}_{it}+{\gamma }_1{Z}_{1it}+{\gamma }_2{Z}_{2it}+{\gamma }_3{Z}_{3it}+{\gamma }_4{Z}_{4it}+e_3\end{array}$$

(4)

where $Y_{it}$ represents the carbon emission intensity of province i in year t, $X_{it}$ represents the information infrastructure construction of province i in year t, and $M_{it}$ represents technological innovation in province i in year t. $Z_{1it}$, $Z_{2it}$, $Z_{3it}$, and $Z_{4it}$ represent a set of control variables. In Equation (2), c represents the total influence effect of information infrastructure construction on carbon emission intensity. In Equation (3), a represents the influence effect of information infrastructure construction on technological innovation. In Equation (4), $c^{\prime }$ represents the direct influence of information infrastructure construction on carbon emission intensity after controlling for technology innovation intermediary variables. b represents the influence of technological innovation on carbon emission intensity after controlling for information infrastructure construction. e is the error term.

4. Results

4.1. Panel Data Correlation Test

The data used in this paper are panel data, and the Hausman test was chosen to determine whether the model was suitable for a fixed effects model or a random effects model. The test results are shown in

Table 4. Hausmann test results.

		Coefficient
	(b)	(B)	(b − B)	sqrt(diag(Vb − VB))
	fe	re	Difference	Std. Err.
IIC	−0.106	0.400	−0.507	0.215
lnTI	−0.593	−0.726	0.133	0.061
lnEnergy	1.294	1.745	−0.450	0.244
lnGDP	0.151	−0.434	0.586	0.300
lnFDI	−0.518	−0.475	−0.042	0.093
lnOpen	0.713	0.617	0.095	0.119
ER	−0.434	−0.367	−0.067	00.039
chi2(7) = (b − B)′[(Vb − VB) − 1](b − B) = 16.46
Prob > chi2 = 0.0212

. According to the p-value obtained from the test, p < 0.05, so at the 5% level of significance, the fixed effects model was chosen as more appropriate.

Next, cross-sectional correlations were tested using three tests: the Pesaran, Friedman, and Frees tests. The results of the tests are shown in

Table 5. Cross-sectional correlation test.

	Pesaran’s CD Test	Friedman Test	Frees Test
Test value	1.458	17.571	3.616
p-value	0.144	0.952	1% threshold: 0.5198

. From the test results, it can be seen that the results of the Pesaran and Friedman tests cannot reject the original hypothesis; that is, the model does not have a cross-sectional correlation problem. The results of the Frees test show that the original hypothesis is rejected at the 1% level of significance, and the model has a cross-sectional correlation problem. Therefore, the Driscoll–Kraay model was used next to deal with the cross-sectional correlation problem.

4.2. Basic Regression Results

First, to control the differences between time and regions, we adopted a dual fixed effects model to explore the relationship between information infrastructure construction and carbon emission intensity. Secondly, as the test model has a cross-sectional correlation, we used the Driscoll–Kraay model to correct for the cross-sectional correlation problem.

Table 6. Basic regression.

Variables	Fixed Effects Model CEI	Fixed Effects Model CEI	Driscoll-Kraay Model CEI
$IIC$	−1.134 **	−0.949 **	−0.949 ***
	(0.441)	(0.425)	(0.371)
$lnOpen$		5.870 ***	5.870 **
		(1.727)	(2.515)
$lnEnergy$		1.792 ***	1.792 ***
		(0.448)	(0.371)
$lnFDI$		−5.636 ***	−5.636 **
		(1.730)	(2.465)
$lnTI$		−0.664 ***	−0.664 ***
		(0.165)	(0.114)
$lnGDP$		3.936 **	3.936
		(1.738)	(2.190)
$ER$		−0.348 **	−0.348
		(0.183)	(0.244)
Constant	−1.303	61.57 ***	61.16 **
	(0.886)	(17.74)	(25.66)
$R^2$	0.948	0.958
Adjusted $R^2$	0.938	0.950
Observations	300	300	300
Fixed year	YES	YES	YES
Fixed province	YES	YES	YES

NOTE: The value in $\left(\right)$ represents the standard error term; “***” and “**” represent significance at the 1% and 5% levels, respectively.

shows the regression results.

According to Equation (1), a double fixed effects model regression was conducted, and the results obtained are shown in Table 6. Column (1) analyses the impact of information infrastructure construction on carbon emission intensity without adding control variables. The coefficient of information infrastructure construction is −1.134, which is significantly negative at the 5% significance level. Column (2) shows the effect of information infrastructure construction on carbon emission intensity after the inclusion of the control variables of openness, technological innovation, energy consumption, level of economic development, environmental regulation, and foreign direct investment. After adding the control variables, the coefficient of carbon emission intensity becomes −0.949, which is significantly negative at a significance level of 5%. Compared to column (1), the value of the information infrastructure construction coefficient is significantly lower, indicating that the information infrastructure construction through the indirect effect of control variables has an impact on carbon emission intensity. As the information infrastructure construction continues to improve, the information and digital technology becomes more refined, enabling companies to transition to low-carbon manufacturing, improving resource utilization and reducing carbon dioxide emissions.

The results in column (3) show that the impact of information infrastructure development on carbon emissions is significant at the 1% level when we deal with cross-sectional-related issues. The resulting coefficients are the same as for the fixed effects model, although the significance has increased.

Throughout columns (1)–(2) of Table 6, the coefficients of the variables in the equation are significant, and the $R^2$ of the fit is gradually increasing, indicating that the explanatory power of the equation is also increasing and further indicating that it makes sense for us to add control variables.

4.3. Intermediary Effect Results

Based on the results of the basic regressions in the previous section, the coefficient on information infrastructure construction decreases significantly when control variables are included. We conjecture that some control variables act as the intermediary effect that influences the impact of information infrastructure construction on carbon emission intensity. Therefore, in order to test our conjecture, we chose the variable of technological innovation in conjunction with the previous theoretical analysis to test the intermediary effect. Common models used to test the intermediary effect include the stepwise regression coefficient method, the Sobel method, and the Bootstrap method. In order to improve the reliability of the test results, we chose two of these methods, namely the stepwise regression coefficient method and the Sobel method.The test results are shown in

Table 7. Intermediary Effects Regression Results.

Variables	(1) CEI	(2) lnTI	(3) CEI
$IIC$	−1.309 ***	0.542 ***	−0.949 ***
	(0.427)	(0.158)	(0.371)
$lnOpen$	5.970 ***	−0.152	5.870 **
	(1.778)	(0.656)	(2.515)
$lnEnergy$	1.277 ***	0.775 ***	1.792 ***
	(0.441)	(0.163)	(0.371)
$lnFDI$	−5.661 ***	0.0386	−5.636 **
	(1.781)	(0.656)	(2.465)
$lnTI$			−0.664 ***
			(0.114)
$lnGDP$	3.400 *	0.805	3.936
	(1.784)	(0.658)	(2.190)
$ER$	−0.223	−0.242 ***	−0.348
	(0.183)	(0.067)	(0.244)
Constant	63.859 ***	−3.44	61.16 **
	(18.251)	(6.736)	(25.66)
Observations	300	300	300
Fixed year	YES	YES	YES
Fixed province	YES	YES	YES

NOTE: The value in $\left(\right)$ represents the standard error term; “***”, “**” and “*” represent significance at the 1%, 5% and 10% levels, respectively.

below.

From the test results, in column (1) of Table 7, when the control variable of technological innovation is excluded, the regression coefficient of information infrastructure construction on carbon emission intensity is significantly negative at the 1% level. This indicates that the construction of information infrastructure reduces carbon emission intensity to some extent, i.e., the test result of coefficient c is significant, and there is a total effect. According to the results in column (2) of Table 7, the regression coefficient of technological innovation on information infrastructure construction is 0.542, which is positive at the significance level of 1%. The results indicate that information infrastructure development can promote technological innovation. In other words, the test result of coefficient a is significant. The results in column (3) of Table 7 show that when technological innovation is added as an intermediary variable, the regression coefficient of information infrastructure construction on carbon emission intensity is −0.949. The value of the coefficient decreases, but it is still significant at the 1% level. That is, the test result of coefficient $c^{\prime }$ is significant, and there is a direct effect of information infrastructure construction on carbon emission intensity after controlling for technological innovation. Looking at the three models as a whole, information infrastructure construction has a direct effect on reducing carbon emissions, and information infrastructure construction promotes technological innovation. After adding the intermediary variable of technological innovation, the coefficient of information infrastructure construction is significantly reduced, which indicates that the mediating variable plays an indirect role. Therefore, technological innovation does have a partial intermediary effect.

Next, the Sobel test was used to test the intermediary effect again to further determine whether technological innovation was an intermediary effect between information infrastructure construction and carbon emission intensity. The test results are shown in the following

Table 8. Sobel test results.

	Coefficient	SD	Z-Statistic	p-Value
Coefficient a	0.542	0.158	3.435	0.001
Coefficient b	−0.664	0.165	−4.029	0.000
Indirect effect a ∗ b	−0.360	0.138	−2.614	0.009
Direct effect c′	−0.949	0.425	−2.234	0.025
Total effect c	−1.309	0.428	−3.062	0.003
Indirect effect as a percentage of total effect: 0.275
The ratio of indirect effects to direct effects: 0.379
Ratio of total effect to direct effect: 1.379

.

According to the results of the Sobel’s method test in Table 4, the p-values corresponding to coefficient a, coefficient b, indirect effect $a\ast b$, total effect c, and direct effect $c^{\prime }$ are all less than 0.05. This indicates that the information infrastructure construction has an impact on carbon emission intensity through the intermediate variable of technological innovation, with an indirect effect of 37.9%, of which the mediating effect accounts for 27.5%.

4.4. Robustness Tests

We next tested whether the results of the regression model mentioned above are robust and avoid the endogeneity problem between information infrastructure construction and carbon emission intensity. In this study, the robustness of the double fixed effects model was tested. The robustness test of the original model was carried out by replacing the explained variables and instrumental variables method to eliminate, as far as possible, any bias in the calculations due to endogeneity issues.

4.4.1. Reconstructing the Dependent Variables

Since both carbon dioxide and sulfur dioxide are produced during energy consumption and both have an impact on air pollution, the sulfur dioxide emission intensity ($SEI$) for 2010–2020 (calculated using the same method as carbon emission intensity) was used to reconstruct the explanatory variables based on data from the China Environment Statistics Yearbook. The following

Table 9. Robustness Tests Results.

Variables	(1) SEI	(2) SEI	(3) 1st Stage	(4) 2nd Stage
$IIC$	−7.010 ***	−4.873 ***		−3.482 ***
	(−1.421)	(−1.411)		(0.921)
L.Z			0.090 ***
			(0.023)
Control variables	NO	YES	YES
Observations	330	330	270	270
Fixed year	YES	YES	YES	YES
Fixed province	YES	YES	YES	YES
LM statistic				6.870
p-value				0.008
Wald F				10.690
KP Wald F				14.610

NOTE: The value in $\left(\right)$ represents the standard error term; “***” represent significance at the 1% levels.

displays the test results.

According to the test results, when no control variables are added in column (1) of Table 9, the emission intensity of sulfur dioxide is significantly negative at the level of 1% with a value of −7.010, which shows that the information infrastructure construction reduces the emission intensity of sulfur dioxide. In column (2), the inclusion of the control variables results in a significantly negative sulfur dioxide emission intensity at the 1% level, with a value of −4.873. The test results indicate that the information infrastructure construction can still reduce the sulfur dioxide emission intensity after replacing the explanatory variables.

4.4.2. Instrumental Variable Regression

There may be endogeneity issues in the basic regression results in Section 4.1. Additionally, there are some non-numerical elements that can also influence carbon emission intensity, such as political factors and cultural factors. This may lead to missing variables or inaccurately estimated coefficients. In addition, high carbon emission intensity cities are likely to be slower to develop technological information, which, in turn, hinders the development of information infrastructure. Therefore, the instrumental variable in this paper is the one-period lag of the historical long-haul fiber optic cable route length. This instrumental variable has a direct effect on the core explanatory variables but not on carbon emissions. Therefore, in this paper, the model is re-estimated using two-stage least squares ($2SLS$) with one cycle lag of the long-haul fiber optic cable line length as the instrumental variable. The test results are presented in columns (3)–(4) of Table 9.

According to the results of the first stage of $2SLS$, the coefficient of the one-period lag of information infrastructure construction is significantly negative at the 1% level with a value of 0.090. The results of the second stage show a coefficient value of −3.482 for information infrastructure construction, which is significant at the 1% level. For the test results of the unidentifiable equation, the value of the $LM$ statistic was 6.870, with a p-value less than 0.05, so the hypothesis that it was unidentifiable was rejected. Based on the test results for the weak instrumental variables, the Wald F statistic was 10.690, and the $KP$ Wald F statistic was 14.61, both of which were greater than the critical value, indicating that there were no weak instrumental variables in the model. Combining all the results shows that information infrastructure construction decreases carbon emission intensity when the underlying endogenous issues are mitigated. The results obtained for the instrumental variables are in line with the model results in the previous paper, which shows that the chosen instrumental variables are robust. This also confirms the research hypothesis in the paper.

5. Heterogeneity Analysis

The empirical results based on data from 30 provinces across China show that information infrastructure construction can significantly reduce carbon emission intensity. To understand the research conducted in more detail and to analyze the differences in impacts under different circumstances, this paper first divides China into different regions for heterogeneity testing. Heterogeneity tests are then conducted separately according to the level of carbon emission intensity and differences in the level of technological innovation to analyze the impact in different contexts.

5.1. Regional Heterogeneity

There are currently many ways of dividing China into regions. Some scholars divide China into regions according to the three economic belts, while others divide China into regions according to geographical location. In this paper, the three economic belts are followed to divide China into the eastern region, the western region, and the central region. The eastern region includes 11 provincial administrative regions, namely Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan. The central region includes eight provincial administrative regions, namely Heilongjiang, Jilin, Shanxi, Anhui, Jiangxi, Henan, Hubei, and Hunan. The western region includes 11 provincial administrative regions, namely Sichuan, Chongqing, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Guangxi, and Inner Mongolia. Tibet is not included in the western region as there are too many missing values in the data. See

Table 10. Results of regional heterogeneity test.

Variables	East (1) CEI	Central (2) CEI	West (3) CEI
$IIC$	−0.948 ***	−4.300 **	1.082
	(0.178)	(1.825)	(0.791)
Control variables	YES	YES	YES
Observations	110	80	110
Fixed year	YES	YES	YES
Fixed province	YES	YES	YES

NOTE: The value in $\left(\right)$ represents the standard error term; “***” and “**” represent significance at the 1% and 5% levels, respectively.

for the results of the regional heterogeneity test.

As can be seen from the test results shown in Table 10, the coefficients of information infrastructure construction in column (1) of the eastern region and column (2) of the central region are significantly negative at the levels of 1% and 5%, respectively, with coefficients of −0.948 and −4.000 for both, which are consistent with our expected results. Numerically, the coefficient of information infrastructure construction in the central region is greater than that in the eastern region. However, the test results for the eastern region are more significant and the model fits better. The coefficient of information infrastructure construction in column (3) for the western region is 1.082, which is not significant, but the fit of the model is relatively good. This may be because western China has a poorer geographical environment and a less developed economy, so the information infrastructure construction started later and developed more slowly than in other regions. At the same time, there are relatively few high-tech industries in western China, so the demand for information infrastructure construction is smaller, which may lead to an insignificant effect of information infrastructure construction on carbon emission intensity.

The regional variability in the impacts on carbon emission intensity is due to the geographical advantage of the eastern and central regions and the help of national policies, resulting in a relatively developed information infrastructure. However, the western region is relatively economically backward, and the construction of information infrastructure is in the process of development. Therefore, the carbon emission intensity is affected differently in the eastern, central, and western regions.

5.2. Heterogeneity in Technology

For each province, differences in the degree of technological innovation may lead to differences in the development of digital industries, changes in industrial structure, and resource utilization. This leads to different impacts of information infrastructure construction on carbon emission intensity. Therefore, technology innovation is divided into low-technology innovation groups and high-technology innovation groups to verify the difference in the influence of information infrastructure construction on carbon emission intensity for different degrees of technology innovation. The results of the test are listed in

Table 11. Heterogeneity results of technological innovation.

Variables	Low-Tech Innovation (1) CEI	High-Tech Innovation (2) CEI
$IIC$	−0.725	−0.602 ***
	(1.339)	(−0.130)
Control variables	YES	YES
Observations	140	160
Fixed year	YES	YES
Fixed province	YES	YES

NOTE: The value in $\left(\right)$ represents the standard error term; “***” represent significant at the 1% levels.

.

According to the test results in Table 11, we can see that the coefficient of information infrastructure construction in column (1) of low-tech innovation is not significant. This is mainly because technological innovation is relatively backward and low, and it is difficult for information infrastructure construction to provide high-quality technical support for its development. Therefore, the information infrastructure construction has little effect in reducing carbon emission intensity. For high-tech innovation in column (2), the coefficient of information infrastructure construction is significantly negative at the level of 1%, with a value of −0.602. This result suggests that the information infrastructure construction significantly reduces the carbon emission intensity of provinces in the high-technology innovation group. This may be due to the fact that better technological innovation accelerates the construction of information infrastructure. Higher technological innovation can support the development of more green and low-carbon industries and, at the same time, improve the resource utilization efficiency of enterprises, thus reducing carbon dioxide emissions.

5.3. Heterogeneity in Carbon Emission Intensity

We will further analyze the impact of information infrastructure development on carbon emission intensity differences. The 30 provinces in China are divided into low-carbon emission intensity groups and high-carbon emission intensity groups. The effects of information infrastructure construction on reducing carbon emission intensity were evaluated separately. The test results can be found in

Table 12. Heterogeneity of carbon emission intensity results.

Variables	Low Carbon Emission (1) CEI	High Carbon Emission (2) CEI
$IIC$	−0.840 ***	−2.551 *
	(−0.175)	(−1.580)
Control variables	YES	YES
Observations	200	100
Fixed year	YES	YES
Fixed province	YES	YES

NOTE: The value in $\left(\right)$ represents the standard error term, “***” and “*” represent significant at the 1% and 10% levels.

.

According to the test results in Table 12, it can be seen that the coefficient of information infrastructure construction in column (1) for the low-carbon emissions group is significantly negative at the level of 1%, with a value of −0.840. In column (2), for the high-carbon emissions group, the coefficient of information infrastructure construction is significantly negative at the 10% level with a value of −2.551, which is significantly larger than that of the low-carbon emissions group. This indicates that the construction of information infrastructure has a significant impact on carbon emissions at different intensities, and the impact is more pronounced in areas with high carbon emission intensity than in areas with low carbon emission intensity. This suggests that information infrastructure is more resilient in high-carbon areas than in low-carbon areas.

6. Discussion and Conclusions

This paper investigates the effect of information infrastructure construction on carbon emission intensity, which is of great practical importance. Based on research data from 30 Chinese provinces from 2010 to 2019, an index of information infrastructure construction was constructed using the entropy method, and then the fixed effects model, Driscoll–Kraay model, and the mediating effects model were used to explore the impact and mechanism of action of information infrastructure construction on carbon emission intensity. Based on this, a heterogeneity analysis was conducted. The conclusions are as follows: (1) Information infrastructure construction has a mitigating effect on carbon emission intensity; (2) Technological innovation exists as an intermediary variable in the impact of information infrastructure construction on carbon emissions, i.e., technological innovation promotes information infrastructure construction to decrease carbon emissions; (3) The impact of information infrastructure construction on carbon emission intensity is heterogeneous. Firstly, there are regional differences, with significant effects in the central and eastern parts of the country but not in the west. Secondly, there are differences in technological innovation, i.e., the impact of information infrastructure construction on carbon emissions is more significant for high-tech innovation groups. Finally, there is a difference in carbon emission intensity, i.e., among high-carbon emission regions, information infrastructure development has a greater impact on reducing carbon emission intensity.

The research presented in this study offers some theoretical justification for how the development of information infrastructure influences carbon emission intensity, as well as some recommendations for national policymakers, which are outlined in the following. (1) Strengthen the development of information infrastructure construction, improve the information infrastructure construction, and raise the level of information infrastructure construction. Firstly, the financial allocation and construction investment for the information infrastructure construction should be increased. Secondly, it is necessary to optimize traditional basic information construction as well as strengthen the construction of new information infrastructure. Resources should be used efficiently to avoid excessive investments and duplication of construction. (2) Encourage technological innovation and development. First, increase funding for technological innovation and encourage enterprises to conduct independent research on and development of low-carbon technologies. Second, improve talent policies and strengthen the training of innovative talents. Third, increase the promotion and use of low-carbon technology innovation, narrow the regional gap in technology innovation, and strengthen resource complementation, $R\&D$ cooperation, and talent exchange between regions. (3) Build efficient information infrastructure according to regional characteristics. In the western region, where the information infrastructure construction is not very well developed, the scale of information infrastructure construction should be expanded. Taking the government as the leader, we should expand investment in the construction of information infrastructure, broaden the investment channels, and attract investors. For the central and eastern regions, where the information infrastructure construction is relatively developed, the construction of high-end information infrastructure should be vigorously developed on the premise of continuously improving the construction of information infrastructure. This will improve the impact of information infrastructure construction on carbon emission reduction and promote the development of low-carbon industries.

The results obtained from this paper can be compared with those of existing studies to reveal several differences. Firstly, the impact of technological innovation as a mediating variable in this paper can enrich the study of relevant carbon emissions. Secondly, the test of heterogeneity by province can reflect regional differences more clearly and allow for better policy formulation.

However, there are still some limitations to the current work. Firstly, the sample is relatively simple and limited to the provincial level of a single country, and the data cannot be updated to the most recent year for the time being. Secondly, only the unidirectional impact of information infrastructure development on carbon emissions is considered in this paper, which is not very comprehensive and the direct bidirectional impact of both could be further explored. Secondly, the environmental problems arising from air pollution are not only due to excessive carbon emissions, and further research could include a more comprehensive range of polluting gases and dust categories. Finally, the choice of control variables may not be comprehensive enough, and a more extensive reading of the literature could be undertaken to include a more comprehensive range of other variables for study.