Relationship between the digital economy, resource allocation and corporate carbon emission intensity: new evidence from listed Chinese companies

In this study, the impact of the digital economy on carbon emissions was analyzed using the borderless organization theory (Ashkenas et al 2015), which states that information, resources, ideas, and energy can quickly cross the boundaries between enterprises, allowing managers to swiftly react to environmental changes. Digital technologies can break down the geographical and knowledge boundaries between firms (Hanelt et al 2021) and make it easier to access information and capital, which can affect carbon emissions (Zhao et al 2021). Omri and Hadj (2020) and Usman et al (2021) found that ICT facilitates knowledge flow between firms to reduce their environmental impact. Lange et al (2020) found that ICT reduced energy consumption through increased energy efficiency and associated sectoral reforms.

Research on the digital economy and carbon emissions has mainly been conducted at the macro level (Khan et al 2020). For example, using panel data from 30 Chinese provinces between 2006 and 2017, Ma et al (2022) found that a larger digital economy led to lower provincial carbon emissions. Investment in R&D not only curbs carbon emissions but also has a moderating effect. Ren et al (2021) constructed a provincial internet development index for China using the entropy weight method and found that internet development reduced energy consumption and associated carbon emissions by changing the energy structure, economic growth, and investment in R&D and human capital.

Based on the above literature, we propose a hypothesis (H1; figure 1): The digital economy reduces corporate carbon emissions.

Zoom In Zoom Out Reset image size

According to the resource-based view, firms have heterogeneous resources, which results in different resource allocation pathways (Wernerfelt 1984). The digital economy is driven by the adoption of numerous digital technologies, which have disrupted the traditional patterns of resource utilization (Ma et al 2022). Various firms have different preferences for adopting digital technologies (Walter et al 2007), which improve resource allocation and increase energy efficiency (Aral and Weill 2007). Digital technologies also help managers monitor the dynamics between producers, consumers, and suppliers to facilitate the adjustment of their production schedules and achieve an efficient allocation of resources (Kotabe et al 2008). For example, Lu et al (2020) found that digital technologies can promote the flow of knowledge, skills, and capital.

There is an increasing amount of studies being conducted on the relationship between resource allocation and carbon emissions. Based on resource allocation theory (Hsieh and Klenow 2009), resource mismatch affects firms' production performance and environmental governance (Brandt et al 2013). Hao et al (2020) analyzed provincial panel data from 30 Chinese provinces from 2005 to 2016 and found that resource mismatch (capital and labor) can reduce the energy efficiency of a region and its adjacent regions. It follows that more efficient resource allocation leads to greater energy efficiency. Similarly, Du and Li (2021) found that the misallocation of land resources increased pollution emissions from firms because of over-investment and production inefficiency.

Based on the above literature, we proposed the following hypothesis (H2; figure 1): Efficient resource allocation further increases the reduction in carbon emissions associated with the digital economy.

In this study, we used a sample of listed Chinese manufacturing companies from 2011 to 2019. The companies' financial data were obtained from the China Securities Market and Accounting Research Database and Wind databases. Carbon footprint data were obtained from the China Energy Statistical Yearbook and Industrial Statistical Yearbook. City-specific variables were obtained from the China City Statistical Yearbook. Excluding companies with a significant lack of data, we analyzed 10,934 annual datapoints.

3.2.1. Dependent variable: corporate carbon emission intensity(CEI)

According to the Intergovernmental Panel on Climate Change (2006), total carbon emissions(${{\rm{C}}}_{{\rm{j}}}$) is the sum of direct and indirect carbon emissions originating from fossil fuel combustion ($C{F}_{j}$) and industrial processes ($C{P}_{j}$), respectively, as shown in the following equation(Wu et al 2019):

$$\begin{eqnarray}&&{{\rm{C}}}_{{\rm{j}}}={{\rm{CF}}}_{{\rm{j}}}+{{\rm{CP}}}_{{\rm{j}}}\end{eqnarray} \tag{ 1 }$$

Carbon emissions originating from fossil fuel combustion in each sector were calculated according to the methodology of Shan et al (2018) using the following equation:

$$\begin{eqnarray}&&{{\rm{CF}}}_{{\rm{j}}}=\displaystyle \sum {{\rm{AD}}}_{{\rm{ej}}}\times {{\rm{NCV}}}_{{\rm{e}}}\times {{\rm{CC}}}_{{\rm{e}}}\times {{\rm{O}}}_{{\rm{ej}}}\times 44/12\end{eqnarray} \tag{ 2 }$$

where $C{F}_{j}$ represents carbon emissions from the consumption of all types of energy in a given sector j (Mt CO₂); $A{D}_{ej}$ denotes the consumption of energy type e by the corresponding sector j (t); $NC{V}_{e}$ is the net calorific value of energy type e (pJ/${10}^{4}$ t); $C{C}_{e}$ represents carbon emissions per unit of net calorific value (ton C/TJ); and ${O}_{ej}$ is the rate of carbon oxidation (unit of measurement).

Carbon emissions originating from industrial production processes include carbon indirectly produced by various reactions (physical or chemical). They were calculated using the following equation:

$$\begin{eqnarray}&&{{\rm{CP}}}_{{\rm{j}}}={{\rm{P}}}_{{\rm{mj}}}\times {{\rm{EF}}}_{{\rm{mj}}}\end{eqnarray} \tag{ 3 }$$

where $C{P}_{j}$ is the sum of indirect carbon emissions associated with sector j; ${P}_{mj}$ is the production of good m by sector j (unit of measurement); and $E{F}_{mj}$ is the carbon emission factor of sector j in the production of good m (Wei et al 2022).

We subsequently calculated sub-sectoral carbon emissions at the firm level based on the approach of Ren et al (2022) using the following equation:

$$\begin{eqnarray}&&C{E}_{i}={C}_{j}\times {O}_{i}/{O}_{j}\end{eqnarray} \tag{ 4 }$$

where $C{E}_{i}$ represents the carbon emissions of corporation i; ${C}_{j}$ denotes the total carbon emissions of the corresponding sector j; ${O}_{i}$ denotes the operating costs of corporation i; and ${O}_{j}$ is total cost of business of sector j (Shen and Huang 2019).

Lastly, we applied the traditional method of measuring carbon emission intensity at the macro level to the firm level (Zhang et al 2020) using the following equation:

$$\begin{eqnarray}&&{{\rm{CEI}}}_{{\rm{it}}}={{\rm{CE}}}_{{\rm{it}}}/{{\rm{OI}}}_{{\rm{it}}}\end{eqnarray} \tag{ 5 }$$

where $CE{I}_{it}$ is the carbon emission intensity of corporation i in year t; $C{E}_{it}$ denotes the carbon emissions of corporation i; and $O{I}_{it}$ is the operating income of corporation i.

3.2.2. Independent variable: digital economy

3.2.3. Mediation variables: resource allocation(RA)

3.2.4. Control variable

We used the following double fixed-effect model to determine the effect of the digital economy on carbon emissions:

$$\begin{eqnarray}&&CE{I}_{i,t}={\alpha }_{0}+\beta D{E}_{i,t}+\displaystyle \sum _{i=1}^{6}{\gamma }_{i}control+{\alpha }_{i}+{\delta }_{t}+{{\epsilon }}_{i,t}\end{eqnarray} \tag{ 6 }$$

where CEI is the carbon emission intensity. DE is the digital economy. Control is control variables. ${\alpha }_{i}$ and ${\delta }_{t}$ are vectors of a firm i and year t, respectively, employed as dummy variables to minimize the influence of un-observable factors on regression results (Lee and Nelder 2009).

To test the mediating role of resource allocation in the relationship between the digital economy and carbon emissions, the following model was developed.

$$\begin{eqnarray}&&R{A}_{i,t}={\alpha }_{0}+\beta D{E}_{i,t}+\displaystyle \sum _{i=1}^{6}{\gamma }_{i}control+{\alpha }_{i}+{\delta }_{t}+{{\epsilon }}_{i,t}\end{eqnarray} \tag{ 7 }$$

$$\begin{eqnarray}&&CE{I}_{i,t}={\alpha }_{0}+{\beta }_{0}D{E}_{i,t}+{\beta }_{1}R{A}_{i,t}+\displaystyle \sum _{i=1}^{6}{\gamma }_{i}control+{\alpha }_{i}+{\delta }_{t}+{{\epsilon }}_{i,t}\end{eqnarray} \tag{ 8 }$$

Table 1 shows the statistical summary of each variable. The mean, minimum, and maximum values of carbon emission intensity were 0.211, 0.002 and 4.357, respectively. This indicates that carbon emission intensity differs between enterprises, the majority of which had a low corresponding value. The mean, minimum, and maximum values of the digital economy were 10.053, 6.092 and 12.803, respectively, indicating that although the digital economy is unequally developed across China, most regions are at a higher level of development. The multicollinearity test found that the mean VIF value was 1.96, which is less than 10 and indicates that there is no multicollinearity.

Table 2 shows the results of the baseline regression with the stepwise addition of control variables. It is evident that the digital economy significantly reduced corporate carbon emission intensity. As seen in column 7, the coefficient of the digital economy was −0.0454 at the 1% significance level. This means that for every 1% increase in the digital economy, corporate carbon emission intensity decreased by 0.0454%. This supports hypothesis H1.

Figure 2 shows the pathways of action of the digital economy. The impact of the digital economy on resource misallocation was −0.008 at the 1% significant level. The effect of resource misallocation on the carbon emission intensity of firms was 0.148 at the 1% significant level. The direct effect of the digital economy on the carbon emission intensity of firms was −0.045 at the 1% significant level. Therefore, resource misallocation increased carbon emission intensity, thereby partially undermining the reduction in carbon emission intensity caused by the digital economy. As determined using the Sobel test $\left(-0.008\times 0.148\right),$ the indirect effect of the digital economy was −0.00125 (p < 0.001). Lastly, as determined using 1,000 bootstraps, the stability of the mediation model was −0.00118 (p < 0.001). This is consistent with hypothesis H2.

Zoom In Zoom Out Reset image size

In table 3, columns 1 and 2 show the regression results after replacing the original dependent variable (carbon emission intensity) with the natural logarithm of carbon emissions as the new dependent variable. Columns 3 and 4 show the robustness of the model after excluding provincial capitals, which receive more policy support (Liu et al 2020). Column 5 shows the regression results obtained using the instrumental variables approach to determine endogeneity. To that end, we used the number of post offices per million people in each city as the instrumental variable (Ma and Zhu 2022). Columns 6 and 7 show the regression results obtained using the two-stage method to test for selection bias. Firms may consider their own corporate characteristics and city-related factors to determine their place of incorporation, which can lead to selection bias. Table 3 shows that the instrumental variables selected were valid and that there was selection bias. Using these four methods, we found that the regression results were plausible.

Considering that the digital economy is affected by both government regulation and market forces, we introduced two variables: government intervention and market drivers. We used city-level fiscal expenditure as a percentage of regional GDP to measure government intervention (Huang and Du 2017). We developed a city-level marketization index evaluation system based on a study by He and Wu (2017), which was then compared with firm data.

In table 4, columns 1 and 2 show the regression results after incorporating government intervention. The coefficient of the digital economy was −0.007 at the 1% significance level and the coefficient of the quadratic term ($\mathrm{digital}\,\mathrm{economy}\times \mathrm{government}\,\mathrm{internvention}$) was 0.001 at the 1% significance level. This means that government intervention decreases the ability of the digital economy to reduce resource misallocation and, by extension, carbon emission intensity. Columns 3 and 4 show the regression results after the inclusion of market drivers. The coefficient of the digital economy was −0.007 at the 1% significance level and the coefficient of the quadratic term ($\mathrm{digital}\,\mathrm{economy}\times \mathrm{market}\,\mathrm{drivers}$) was −0.0003 at the 10% significance level. This implies that the market improves the ability of the digital economy to reduce resource misallocation and, by extension, carbon emission intensity. Lastly, columns 5 and 6 show the regression results obtained when both government intervention and market drivers were considered, which was consistent with the previous findings.

Table 4. Moderating role of government intervention and market driven.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	RA	CEI	RA	CEI	RA	CEI
RA		0.144 a		0.147 a		0.142 a
		(0.024)		(0.024)		(0.024)
DE	−0.007 a	−0.035 a	−0.007 a	−0.037 a	−0.007 a	−0.031 a
	(0.001)	(0.004)	(0.002)	(0.004)	(0.002)	(0.004)
${\rm{DE}}\times \,$Goi	0.001 b	0.008 a			0.001 c	0.005 a
	(0.0004)	(0.001)			(0.0005)	(0.001)
${\rm{DE}}\times \,$Mad			−0.0003 c	−0.003 a	−0.001 c	−0.027 a
			(0.0004)	(0.001)	(0.001)	(0.003)
${\rm{DE}}\times {\rm{Goi}}\,\times \,$Mad					−0.0008	−0.015 a
					(0.001)	(0.002)
C	0.827 a	0.555 a	0.815 a	0.437 a	0.823 a	0.439 a
	(0.022)	(0.058)	(0.024)	(0.064)	(0.026)	(0.067)
Firm	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes
Obs	10,922	10,922	10,926	10,926	10,922	10,922
R-sq	0.084	0.606	0.083	0.605	0.084	0.609

^a indicates significance at the p < 0.01 ^b indicates significance at the p < 0.05 ^c indicates significance at the p < 0.1.

This study makes several contributions to the literature on digital economy and carbon emissions. First, our results indicate that the digital economy can significantly reduce corporate carbon emission intensity at the micro level. This is consistent with the findings of Ma et al (2022), who demonstrated that the digital economy can reduce urban carbon emissions. One possible explanation for this is that the digital economy may help traditional industries to access external resources, which can be used to transform processes and upgrade technology in ways that reduce their carbon intensity. Alternatively, the digital economy often changes traditional business models in a way that reduces the waste of non-essential resources, thereby decreasing carbon emissions linked to production and services. Due to one or both of these processes, these results provide strong evidence for policymakers and business managers.

Second, this study shows that resource allocation has a significant impact on the relationship between the digital economy and corporate carbon emission intensity. Specifically, the digital economy improves resource allocation and resource utilization efficiency, thereby decreasing carbon emission intensity. The use of digital technologies, such as big data, the Internet of Things, and robotics can help business managers to monitor real-time changes related to product production, sales, and supply, which would facilitate the adaptation of their strategies for improved resource utilization efficiency (Ulas 2019). Furthermore, digital technology can promote technological innovation, which improves resource allocation efficiency (Brissimis et al 2010). In contrast, higher degrees of resource mismatch corresponded to higher corporate carbon emission intensity. One possible explanation for this is that resource misallocation is linked to the over-allocation of labor and under-allocation of capital, which reflects distortions in factor markets (Jefferson et al 2006). Such distortions can affect industrial structures and technological upgrades, which in turn affect energy efficiency (Lv et al 2017).

Third, we found that market driver and government intervention improved and hindered the ability of the digital economy to decrease corporate carbon emission intensity, respectively. This is consistent with the findings of most studies (Joo et al 2018, Zhao et al 2018). Government intervention often works against the profit goal of firms in favor of political objectives, which can lead to inefficient resource allocation (Chen et al 2011). The influence of market drivers is linked to the free movement of market factors (Leys 2001). This allows firms to break down organizational boundaries and gain access to external support, such as capital and technology, to improve resource allocation efficiency and thus reduce carbon emissions (Pundziene et al 2021).