How does the opening of high-speed rail drive energy restructuring? New micro evidence from China

The utilization of fossil fuels exerts a pivotal influence in the industrial production of human society, and the global geopolitical landscape exerts a substantial influence on the dynamics of the global energy market (Antonakakis et al. 2017; Ben Cheikh and Ben Zaied 2023). The Russia-Ukraine conflict is a turning point in world history that has had a profound effect on the international system, the global order, and the natural world (Xin and Zhang 2023; Solokha et al. 2023; Pata et al. 2023; Chishti et al. 2023). To diminish their reliance on Russian energy, major energy-consuming nations across the globe have initiated changes in the composition of their energy consumption, progressively elevating the radio of green energy while concurrently reducing reliance on fossil fuels (Abay et al. 2022; Ha 2023; Colgan et al. 2023). On the other hand, traditional fossil fuels, which are the primary source of energy consumption, cause a variety of environmental pollution issues, and energy conservation and emission reduction have become a universal strategy (Adebayo et al. 2020; Depren et al. 2022). Typically in China, a rapidly developing country, the swift economic expansion has led to a heightened consumption of fossil fuels. In 2020 General Secretary Xi Jinping pledged to stop China's carbon dioxide emissions from growing by 2030. Therefore, improving energy efficiency, reducing pollution emissions, and realizing a green economy are of great practical significance to China and the world. Transportation infrastructures have important positive significance for the development of technology and industrial structure (Gong et al. 2023; Liu et al. 2022; Lumeng et al. 2023), and enhancing efficiency and service flexibility can contribute to a decrease in energy consumption and overall costs within the transportation system (Guo et al. 2016). As of 2023, the operational mileage of high-speed railways in China is projected to reach 42,000 km, this milestone positions China as the country with the most extensive high-speed railways operational mileage globally (Huang and Zong 2020). Certainly, the effect of HSR on the progress of industries and technologies with the aim of reducing energy consumption is a question that warrants investigation and clarification. In the realm of infrastructure research, existing literature predominantly concentrate on investigating the effects of HSR on economic efficiency and eco-efficiency (Li and Cheng 2022; Guo et al. 2020; Li et al. 2023). In contrast, empirical knowledge to study how HSR affects energy consumption is still limited. In consideration of this, we consider the implementation of HSR in China as a quasi-natural experiment and analyze its influence on the environment. In detail, we adopt the difference-in-differences model (DID) and difference-in-difference-difference model (DDD) to evaluate the effect of HSR on energy consumption, and then we utilize PSM-DID, placebo test, and other robustness tests to verify the highlights. To our best knowledge, this paper may provide three folds marginal contributions to existing literature: firstly, our primary academic contribution lies in elucidating the mechanisms through which the initiation of high-speed rail influences energy restructuring. Furthermore, we aim to investigate whether, and if so, how HSR affects the energy mix, providing insights to inform sustainable growth policies and strategies. Secondly, existing research mostly pays close attention to exploring macro level impact of HSR operation on human capital flows and regional economic expansion. This paper takes the study to the firm level, using industrial firm data and industrial firm pollution data, and dissects the intrinsic mechanism of HSR's impact on energy restructuring based on the micro level. From the aspect of the effect of HSR opening on enterprises, it is verified that HSR achieves a change in energy consumption structure through the technological upgrading effect and industrial restructuring effect. Thirdly, this paper extends research on the energy rebound impact at the firm level within China's industrial sector, utilizing data spanning from 2003 to 2012. The objective is to offer an objective foundation for policy-making departments to formulate sensible and effective energy policies. In summary, this study holds significant value for the research on the sustainable growth impacts of HSR in China, which offers policy recommendations and strategic implementation for the government. The subsequent sections are organized as follows. Section 2 provides the literature review and the methodology and empirical data are shown in Section 3. The detailed analysis and mechanism analysis are represented in Section 4 and Section 5. Finally, the crucial conclusions and policy implications are draw in Section 6.

The study situates itself within infrastructure and environmental economics literature. Prior research largely examines HSR impacts on economic efficiency and eco-efficiency (Li and Cheng 2022; Guo et al. 2020; Li et al. 2023), with limited empirical evidence on energy consumption structure. The paper draws on studies linking technological innovation, industrial structure change, and government intervention to enterprise energy use (Chen et al. 2020; Fu 2018; Liu et al. 2021; Lu and Zhang 2022; Zhou et al. 2022). It proposes that HSR affects energy use via: (a) a technological innovation effect—HSR reduces spatial frictions, enhances knowledge spillovers, expands markets and profits, and incentivizes R&D and adoption of energy-efficient technologies; and (b) an industrial structure effect—HSR alters labor costs and spatial allocation, encourages relocation of highly polluting firms, and fosters agglomeration of services and high-tech industries with lower pollution intensity. The paper also acknowledges a possible energy rebound effect whereby efficiency gains may initially increase energy use (Berkhout et al. 2000). Based on this, it advances two hypotheses: H1: HSR optimizes the energy consumption structure; H2: this optimization mainly operates through technological innovation and industrial restructuring.

Research design: The opening of HSR in Chinese cities is treated as a quasi-natural experiment. The authors estimate multi-period difference-in-differences (DID) models to identify the effect of HSR on firm-level fossil energy consumption, and augment identification with difference-in-difference-in-differences (DDD) models to capture heterogeneous impacts across enterprise, industry, and regional groups. Models: - Baseline DID: Y_it = β HSR_it + X_it'δ + μ_i + λ_t + θ_c + ε_it, where Y_it is firm i's log consumption of fuel coal (Incoal) or fuel oil (Inoil); HSR_it is a dummy equal to 1 from the year a city's HSR opens onward; X_it are nine firm-level controls; μ_i, λ_t, and θ_c denote firm, year, and city fixed effects. Additional high-dimensional fixed effects are included per specifications (e.g., city-year, industry-year, ownership-year). - DDD: Interacts HSR with grouping dummies (group ∈ {eci high-energy-consuming industry, tech technology-intensive, labor labor-intensive, cap capital-intensive, city small/medium city, np firm produces new product}) to estimate differential effects: Y_it = β1 HSR_it×group + β2 HSR_it + β3 time×group + β4 treat×group + controls and fixed effects, where treat=1 for cities that opened HSR during 2003–2012; time=1 from the opening year onward. Data and variables: - Data sources: HSR openings from China Railway Yearbook and 12306; firm characteristics from China Industrial Enterprise Database and Corporate Green Development Database. Matching via firm identifiers (business code/name, postal code, location). Sample spans 2003–2012 with 477,568 matched observations overall; main regressions use relevant subsamples (e.g., 185,511 obs for coal consumption; 12,136 for oil in Table 1). - Outcomes: Log fuel coal consumption (Incoal) and log fuel oil consumption (Inoil). - Treatment: HSR_it dummy equals 1 for a city from its HSR opening year onward. - Grouping variables (DDD): eci (high-energy-consuming industries per 2010 National Economic and Social Development Statistics Report), tech, labor, cap, city (small/medium city with resident population <1 million), np (produces new products), innovation pilot city, and resource-based city. - Controls: nine firm-level covariates—age; scale (ln total assets); employment (ln employees); long-term liabilities to total assets; current liabilities to current assets (liquidity index); debt-to-asset ratio (solvency); interest expense to debt (financial costs); administrative expenses to total profit (G&A); profitability (sales profit to sales revenue). Identification and robustness: - Parallel trends: Event-study specification with leads/lags (k from -4 to +4) relative to opening year (k=0). Patterns indicate an initial rebound effect followed by significant reductions in fossil energy use two or more years post-opening. - Placebo: Randomly assign HSR openings to cities (500 iterations) and re-estimate baseline; coefficients center around zero with most p-values >0.1, while actual HSR estimates fall in the tails, supporting causality. - PSM-DID: Estimate propensity scores via logit using the nine covariates; 1:2 nearest-neighbor caliper matching (0.01) to select control cities; re-run multi-period DID on matched sample; balance checks (kernel density) provided in appendix. - Additional checks: Subsample restrictions excluding never-HSR cities, provincial capitals/sub-provincial cities, and dropping HSR cities three years pre-opening; results robust. Fixed effects and inference: Specifications control for firm, year, city, and high-dimensional interactions (e.g., city-year, industry-year, ownership-year as noted). Standard errors are robust and clustered at the firm level.

- Baseline DID (Table 1): HSR opening significantly reduces fuel coal consumption at firms (Incoal: β = -0.075, t = -2.89, p<0.01). The effect on fuel oil is negative but not statistically significant in the base model (Inoil: β = -0.067, t = -0.47). - DDD for high-energy-consuming industries (Table 1, cols 3–4): Stronger reductions for high-carbon industries (Incoal: β = -0.399, t = -14.38, p<0.01; Inoil: β = -0.570, t = -4.23, p<0.01). - PSM-DID (Table 2): After propensity score matching, HSR still significantly reduces fossil fuel consumption; e.g., Incoal β = -0.236 (t = -21.48, p<0.01). Subsample and alternative specifications remain consistent; oil results become significantly negative in matched samples (e.g., Inoil β = -1.036, t = -3.63 for certain subsamples). - Parallel trends and rebound: Event-study indicates a short-term rebound around the opening year, with significant and persistent reductions in fossil fuel use emerging two or more years post-opening. - Heterogeneity across enterprise characteristics (Table 3, Panel A): • Innovation (HSR_np): Firms producing new products show larger reductions in coal consumption (Incoal: β = -0.161, t = -3.27, p<0.01); oil effect not significant. • Ownership (HSR_pro): No significant differences by ownership type. • High-carbon firms (HSR_highcar): Additional coal reductions (Incoal: β = -0.055, t = -1.99, p<0.05). - Heterogeneity across industry characteristics (Table 3, Panel B): • Technology-intensive (HSR_tech): Significant reductions in coal (Incoal: β = -0.099, t = -2.72, p<0.01); oil effect not significant. • Capital-intensive (HSR_cap): Significant reductions in oil (Inoil: β = -0.200, t = -1.87, p<0.10); coal effect not significant. • Labor-intensive (HSR_labor): Oil consumption increases (Inoil: β = 0.161, t = 1.81, p<0.10); coal effect not significant. - Heterogeneity across regional characteristics (Table 3, Panel C): • Innovation pilot cities (HSR_polit): Significant coal reductions (Incoal: β = -0.144, t = -3.80, p<0.01); oil effect not significant. • Small/medium cities (HSR_size): Significant coal reductions (Incoal: β = -0.081, t = -1.71, p<0.10). • Resource-based cities (HSR_type): Significant oil reductions (Inoil: β = -0.802, t = -2.56, p<0.05); coal effect not significant. - Mechanisms—Technological progress (Table 4): HSR increases labor productivity (Inoutput and Insales: β = 0.020, t ≈ 6.2–6.3, p<0.01). Convergence effect: firms already producing new products experience smaller productivity gains (HSR_np: β ≈ -0.026 to -0.025, p<0.01). High-carbon industries gain more productivity post-HSR (HSR_highcar: β ≈ 0.019, p<0.01). - Mechanisms—Industrial restructuring (Table 5): HSR increases firm entry and exit (fentry: β = 0.015, t = 22.20, p<0.01; fout: β = 0.011, t = 15.88, p<0.01). For high-carbon industries, both entry and especially exit rise (indentry: β = 0.008, t = 2.63, p<0.01; indout: β = 0.037, t = 24.60, p<0.01), indicating active reallocation and upgrading. - Overall: HSR opening robustly reduces firms’ fossil fuel use—especially coal—via enhanced innovation and industrial upgrading, with stronger effects in high-carbon industries, technology/capital-intensive sectors, small/medium and innovation pilot cities, and resource-based cities.

The study addresses whether and how HSR opening reshapes firms’ energy consumption. Treating HSR rollout as a quasi-natural experiment reveals that HSR reduces fossil fuel consumption, particularly coal, after accounting for rich fixed effects and firm controls. The observed short-term rebound suggests that initial efficiency gains and production expansion can temporarily lift energy use, but as innovation diffusion and structural shifts take hold, net consumption declines. Mechanism tests support two pathways: (1) a technological innovation channel, where HSR compresses spatial frictions, improves access to talent and knowledge, expands markets, and raises productivity, enabling adoption of cleaner technologies and more efficient processes; and (2) an industrial restructuring channel, where HSR-induced changes in accessibility and costs prompt firm relocation, higher entry and exit, and a shift toward services and high-tech industries with lower fossil intensity. Heterogeneity analyses indicate that benefits are concentrated among high-carbon, technology- and capital-intensive firms, and in small/medium, innovation pilot, and resource-based cities. The positive oil response in labor-intensive industries highlights potential trade-offs requiring targeted policy to prevent fuel-switching or unintended increases in oil use.

This paper provides micro-level evidence that the opening of high-speed rail in China significantly optimizes firms’ energy consumption structure, notably reducing coal use and, under matched and subgroup analyses, lowering oil use as well. By exploiting a multi-period DID and DDD framework with extensive robustness checks, it shows that HSR’s effects operate primarily through technological upgrading (productivity gains, innovation diffusion) and industrial restructuring (reallocation via higher entry/exit, sectoral upgrading). The low-energy-consumption effects are heterogeneous, being stronger in high-carbon industries, technology- and capital-intensive firms, small and medium-sized cities, innovation pilot cities, and resource-based cities. Policy implications include: continuing to expand HSR coverage to leverage its non-economic environmental benefits; tailoring strategies by city size and innovation capacity; and promoting ecological protection and industrial upgrading. Policymakers should encourage the growth of technology- and capital-intensive industries and support green technology diffusion, while carefully managing labor-intensive sectors to avoid increases in oil consumption. These insights provide actionable guidance for China and other emerging economies pursuing energy restructuring and low-carbon transitions.