The Effect of Energy Structure on Pollution Emissions: The Role of Technological Innovation

I. Introduction

China’s industry has developed dramatically since 1978. The share of the country’s three major industries in GDP shifted from 27.7%, 47.7%, and 24.6%, respectively, in 1978 to 7.7%, 37.8%, and 54.5% in 2020 (NBSC^[1], 2021). However, the fast expansion of these industries poses great pressure on energy consumption. Specifically, the share of coal consumption in total energy consumption has fallen from 68.5% in 2000 to 57.7% in 2019; this ratio for the industrial sector remains high, with a value of 79.44%, implying that coal-based energy consumption is the main resource for stimulating China’s industrial growth. Unfortunately, total discharge of pollutants currently exceeds the environmental capacity to absorb them (Li et al., 2019).

Economists have contended that technological innovation (TI) is an enabler of economic growth. Schumpeter (1912/1934) incorporated TI into economic analysis and further stressed that economic growth is an evolutionary process with TI as its core. Romer (1990), Grossman & Helpman (1991), and Aghion & Howitt (1992) considered R&D as a form of a firm’s decision-making and endogenised the impact of TI on economic growth. It is agreed that TI is vital for countries worldwide to achieve green development (Shao et al., 2021). Green TI is an essential enabler for facilitating green development (Wurlod & Noailly, 2018).

Two opposite viewpoints can be identified regarding the effect of TI on sustainable development. TI, in particular green technologies, can promote sustainable development (Ghisetti & Quatraro, 2017) by reducing pollutant emissions (PE) (Blum-Kusterer & Hussain, 2001). By developing environmentally-friendly technologies, enterprises can lower energy consumption and limit PEs to ensure cleaner production. Besides, resource recovery technologies can help recycle production wastes, improving energy utilisation efficiency. Green TI facilitates a leap for rising economies from the high-pollution phase of nascent development to the “harmonisation” phase of the environmental Kuznets curve. Considering that TI has opportunity costs, technology transfer with a low rate could make innovation gains less than its opportunity costs, thereby reducing the intensification degree of economic growth. Still, TI may increase energy consumption through the rebound effect (Vélez-Henao et al., 2019). Firms usually neglect environmental costs to maximise their profits (J. Zhang et al., 2018), resulting in the exacerbation of environmental pollution.

It is noteworthy that energy consumption need to be assessed in close conjunction with people’s living standards to understand their net affects (Sethi & Dash, 2022). Because various indicators of quality-of-life index are considered to be highly associated with energy consumption (Mazur, 2011), energy consumption plays a crucial role in the determination of the Human Development Index and the level of sustainable development (Van Tran et al., 2019). Energy consumption is demonstrated by Li et al. (2022) to be the main contributor to environmental deterioration. Zhang et al. (2021) assert that coal-based energy structure (ES) is the main source of greenhouse emissions.

Against this background, it is interesting to question what the role of TI is in the relation between ES and PE reduction at the industry level. Specifically, the study explores whether industrial ES contributes to PEs and how TI affects PEs through ES. The remainder of this paper is structured as follows. Section II designs the methodology. Section III reports the empirical findings, while Section IV makes conclusions.

II. Methodology

Using a panel data of 30 Chinese provinces from 2000 to 2016, we first examine the linear effect of industrial ES on PEs by implementing a dynamic spatial Durbin model (DSDM). We measure ES as the proportion of industrial coal consumption to industrial energy consumption (ES1) and the ratio of industrial energy consumption to business value (ES2), and PE as sulfur dioxide (SO₂) emission per unit of business value (SO2Intensity). The model is as:

$$\begin{align} {lny}_{it} &= {\alpha lnX}_{it} + {\tau lny}_{it - 1} + {\theta wlnX}_{it} \\ &\quad + \mu_{i} + \lambda_{t} + \varepsilon_{it} \tag{1} \end{align}$$

where $y_{it}$ denotes the SO2Intensity; $X_{it}$ represents the regressors, including total population (lnpop), economic growth proxied by per capita gross domestic product (lnrpgdp), average assets of firms (lnaaf), completed investment in wastewater (lniwt), and waste gas treatment (lnwgt); $\alpha$ is the direct coefficient of the explanatory variables; $\theta$ is the spatial lag coefficient of the explanatory variables; $\tau$ is the temporal lag coefficient of the explained variable; $\mu_{i}$ and $\lambda_{t}$ denote individual and time-period fixed effects, respectively, while $\varepsilon_{it}$ denotes the error term. The indicator $w$ is the spatial weight matrix, which is determined by the product of the geographical distance weight matrix and the diagonal matrix of the proportion of real gross domestic product (Yao et al., 2021).

To better deal with the endogeneity problems between variables, we adopt the dynamic panel threshold model (DPTM). This model also allows us to verify the potential nonlinear relationship between ES and SO2Intensity under TI. The DPTM is specified as:

$$\begin{align} ln y_{i t}&=\beta+\alpha_1 ln y_{i t-1}+\alpha_2 ln S O_2 \cdot I\left(\gamma_{i t} \leq \kappa\right)\\ &\quad +\alpha_3 ln S O_2 \cdot I\left(\gamma_{i t}>\kappa\right) \\ &\quad +\sum_{k=1}^5 \rho_k X_{k i t}+\mu_i+\lambda_t+\varepsilon_{i t} \tag{2} \end{align}$$

where $\gamma_{it}$ is the threshold variable (TI); $I( \cdot )$ represents the indicator function; and $\kappa$ is the specific threshold value.

The data mainly comes from the China Statistical Yearbook, China Environmental Statistical Yearbook, and the National Bureau of Statistics of China. The monetary variables have been deflated to 2000 constant prices.

III. Findings

As depicted in Fig. 1, most provinces are mainly distributed in the first and third quadrants, indicating that provinces with higher SO2Intensity are spatially adjacent. In comparison, provinces with lower SO2Intensity tend to be concentrated. The values of global Moran’s I are all significantly positive at a 1% significance level, indicating SO2Intensity has a positive spatial autocorrelation.

Figure 1.Results of Moran scatterplot and global Moran’s I.

Second, the significance of SO2Intensity’s coefficients proves the “accumulation effect” of PE. The coefficient of ES1 in column 2 and its spatial lag term in columns 1 and 2 are positive at a 1% significance level after controlling covariables, suggesting that coal-based ES is a significant factor causing PEs in local and adjacent areas (W. Zhang et al., 2021). Under the constraint of technological capability, the increase in output requires the massive investment of capital and energy, which is associated with dramatical PEs (see columns 4 and 6). Eq. (1) is further extended by adding the interaction term $(ES \times TI)$ to verify the moderating effect of TI, as reported in columns 3 and 6.^[2] Similar to Carrión-Flores & Innes (2010), TI significantly contributes to SO2Intensity. Polluted environments could negatively affect the emotions of corporate executives, causing them to reduce investment in the technology commercialisation process (Lin et al., 2021).^[3] While the collaborative effect between ES and TI inhibits SO2Intensity, the corresponding coefficient in column 6 is statistically significant.

We conduct a robustness test by constructing a new spatial weight matrix using Eq. (3). Table 2 proves the reliability of the benchmark regression. The alternative spatial weight matrix is constructed as follows:

$$w_{i j}^2=\left\{\begin{array}{cc} \frac{1}{d_{i j}}\left(\frac{R G D P}{R G D P}\right)^{\frac{1}{2}} &, i \neq j \\ 0 &, i=j \end{array}\right. \tag{3}$$

where $d_{ij}$ is the geographical distance of province $i$ and $j.$

On the one hand, professional talents and resources will be absorbed by developed regions in a shorter period, which is caused by the “polarisation effect”, thereby curbing TI’s improvement in the adjacent areas. On the other hand, the so-called “trickle-down effect” implies that the local area will feedback surrounding areas through capital outflow and technology overflow, thereby reducing PEs in surrounding areas in the long run.

Third, the above analysis shows that TI negatively moderates local SO2Intensity but has a positive moderating effect in adjacent areas. We infer that the effect of TI may not function as a mutational model but as a structural-gradual model. Under different TI modes, ES may have different effects on SO2Intensity. Subsequently, the two-step system generalized method of moments DPTM is performed by treating TI as a threshold variable.

According to the Wald statistics, all models exhibit significant threshold effects, indicating that the impact of ES on SO2Intensity is nonlinear. Specifically, the Sargan test shows no second-order autocorrelation for the random error term; the Hansen test proves the feasibility of the instrumental variables. Thus, the SPTM regression results reported in this paper is credible. Specifically, when TI exceeds 2.6300, the positive effect of ES1 on SO2Intensity gradually decreases. The “Porter Effect” will force firms to re-articulate production layout, thereby realising “win-win” situation.

By contrast, when TI is higher than 2.5290, the effect of ES2 on SO2Intensity increases monotonically. For emission mitigations and energy conservations, tradeoffs between the economic and environmental goals should be addressed by balancing production adjustment and ecological investment. However, technological enhancements significantly impact ecology but require a large investment in money and time (Xu et al., 2017). Compared to the eastern region, China’s central and western regions lagged far behind regarding economic strength and TI levels. Thus, the increase in total energy consumption is expected to increase pollution emissions in a shorter period because of restriction of TI and other growth-improving activities.

IV. Conclusions

We investigate the industrial ES–PE nexus using the DSDM and DPTM models. Our findings show that ES significantly promotes PE. TI’s improvement can alleviate the restraining effect of coal-based ES on PE. Similarly, industrial energy consumption also significantly increases PE, but this increase occurs in parallel with an increase in the proportion of R&D input.

Our study has some important implications. First, government should firmly carry out supply-side structural reform to force the transformation and adjustment of industries by attracting high-tech industries. Second, as the spatial spillover effect of TI has not yet been established, thus, a sound regional cooperation mechanism should be constructed to better play the spatial spillover effect of TI. Third, sufficient funds from financial institutions and local governments are also required. This is because the innovative capability of the eastern region is better than the central and western regions. In other words, the collaborative effect between ES and TI may exist significant regional heterogeneity, which also provides a crucial information for us further exploration.