The Concept and Classification of ER

The word “regulation” is a legal term that means to regulate and restrict the behavior of economic subjects through laws, regulations, and macro systems. Previous studies have defined “regulation” as “the sum of a series of mandatory regulatory measures enacted and implemented by the state to meet the demands of certain interest groups”. Li and Liu (2023) hold that ER is an environmental standard or action formulated by the government to solve the external problems caused by environmental pollution [20].

ER can be classified into three types [21]: the first is command-and-control ER. These regulations involve government-imposed rules and measures that compulsorily restrict pollution discharge. This includes establishing environmental criteria for pollution emissions or product manufacture, as well as administrative penalties for companies that exceed pollution emission regulations. The second is a market-based incentive ER. These regulations use market mechanisms to guide enterprises in reducing pollution emissions. Examples include pollution charge systems and tradable permit systems. The third is public-participation ER. These regulations rely on environmental petitions, media pressure, and other forms of social opinion to compel the government to strengthen environmental supervision or companies to reduce pollution emissions. This includes public participation in environmental matters, transparency of environmental data, and voluntary environmental agreements [20]. Among these, formal ER includes command-and-control and market-based incentive ER, whereas public-participation ER is classified as informal ER [4, 22].

The Impact of ER on CI

The present literature focuses on three aspects of ER’s impact on CI. First, some scholars have identified a positive effect of ER on CI. Sinn (2008) introduced the “Green Paradox” effect, suggesting that fossil fuel producers, anticipating future declines in demand due to ER, might increase their current extraction rates, thus promoting carbon emissions [23]. Subsequent scholars empirically tested the positive impact of ER on CI. Wang and Wei (2020) discovered that tight ER adoption can have a significant “Green Paradox” effect on economic progress [24]. Zhang et al. (2021) identified this effect in China, where competitive ER behaviors among regions lead to pollution transfers, increasing carbon emissions [25]. Yin et al. (2022) noted that at lower ER intensities, the “Green Paradox” is prevalent, shifting to a carbon reduction effect with stronger ER [2]. Huang and Tian (2023) found that formal ER can inhibit Pat before peak emissions, thus generating a “Green Paradox” effect [4]. Additionally, Sorrell (2009) discussed the “Jevons Paradox”, where strict ER prompts companies to enhance energy efficiency through technological advancements, which paradoxically leads to more energy demand and pollution emissions [26].

Second, numerous scholars have identified a negative impact of ER on CI, highlighting the “forced emission reduction” effect. This effect arises as local governments continually elevate ER standards, compelling high-energy-consuming and high-polluting enterprises (HHE) to either innovate technologically or upgrade pollution control technologies, thereby promoting emission reductions. For example, Hashmi and Alam (2019) found that OECD countries’ CI can be decreased by raising environmental fees and awarding more ecologically favorable patents [27]. Zhang et al. (2022) indicated that the current carbon emission trading system has suppressed green Pat but significantly reduced CI [28]. Chang et al. (2023) observed that ER is positive for CO₂ emission reduction by promoting green innovation [29]. Investment-based environmental regulation is the most effective, followed by command-and-control regulation. Li et al. (2023) demonstrated that ER can reduce CI in regions highly dependent on natural resources [30].

Third, some scholars have identified the dynamic characteristics of ER and CI. Guo and Chen (2018) noted that when ER measures are supported, China’s economic reliance on fossil fuels shifts from weak to strong and back again [8]. The relationship between ER and carbon intensity has a noticeable inverted “U” shape. Wang and Zhang (2022) discovered that the direct influence of ER on carbon emissions similarly has an inverted “U” shape [11]. ER can successfully reduce CO₂ emissions in Chinese cities by encouraging industry structure improvement and technological innovation. Huang and Tian (2023) and Yang et al. (2023) demonstrated the inverted “U”-shaped impact that both formal and informal ER have on net CO₂ emissions [4, 31].

Regional Disparities and Spatial Interaction of ER on CI

Some scholars have begun to focus on the regional differences in the impact of ER on CI. The influence of ER policies on CI varies significantly by region [9, 32]. Industrial investment is limited in regions with stricter ER, whereas industrial investment institutions thrive in regions with relatively weaker ER [32]. Guo and Chen (2018) discovered that the moderating influence of ER on CI was stronger in eastern than in middle and western China [8].

Furthermore, geographical links between environmental regulation and carbon emissions are receiving attention. Cheng et al. (2017) found that CO₂ emissions exhibit significant local spatial clustering [33]. Command-and-control ER effectively reduces emissions but does not significantly impact technological progress, whereas market-based ER promotes technological advancement but is less effective in reducing emissions. Some studies have employed the SDM to examine the impact of ER on CI. Zhang et al. (2021) observed that competition among local governments in China leads to the nearby transfer of pollution and that spatial spillovers of ER are particularly pronounced in the western regions [25]. Huang and Tian (2023) discovered that China’s net CO₂ emissions decrease from east to west and exhibit considerable regional clustering [4].

In summary, the concept of ER continues to evolve with ongoing research. The impact of ER on CI remains a debated topic, necessitating further investigation into ER’s carbon emission reduction effects. In conclusion, the concept of ER has evolved with research. The effect of ER on CI is still a controversial topic, and further research is needed to study the effect of ER on carbon emission reduction. The function of ES in the emission reduction process is clarified by comparing various mechanisms. While some scholars have begun exploring regional differences and spatial connections in the impact of ER on CI, their analyses have largely been static. The dynamic spatial effects of ER on CI require further study. This study seeks to fill this gap by investigating ER’s time and spatiotemporal lag effects on CI from a dynamic spatial viewpoint, assessing direct, indirect, and total impacts in both the short and long term.

The Promoting Effect of ER on CI

First, according to the “Green Paradox” effect, when the government tightens ER, it prompts expectations of even stricter regulations in the future, leading to increased energy costs. Consequently, businesses accelerate their current production activities to mitigate future costs, resulting in a short-term surge in fossil energy demand. This increased energy consumption promotes higher carbon emissions [23]. Therefore, ER may lead to a short-term rise in CI.

Second, according to the “Jevons Paradox” hypothesis, stringent ER can lead companies to improve energy efficiency through technological advancements. However, this increase in efficiency paradoxically stimulates higher energy demand, thereby exacerbating carbon emissions [26]. Consequently, stringent ER may accidentally contribute to increased CI.

The Inhibitory Effect of ER on CI

Two mechanisms, Pat and IS, are also adopted to compare the emission reduction effect of ES.

The first is energy structure. ER primarily influences the ES from both supply and consumption perspectives, and transforming the energy consumption structure can reduce CI [13, 34]. On the supply side, the development of clean energy sources like solar, wind, hydropower, and biomass is encouraged by ER policies, which raise support for renewable energy and enhance the output share of clean energy. On the demand side, ER policies mandate reductions in fossil fuel consumption and encourage a higher proportion of clean energy use, thereby reducing CI. For instance, the “Notice on Doing a Good Job in the Full Coverage of Renewable Energy Green Power Certificates to Promote the Consumption of Renewable Energy Electricity” by the National Development and Reform Commission, the Ministry of Finance, and the National Energy Administration³ proposes strengthening the green electricity responsibility of HHE and increasing their required level of green electricity consumption.

The second is technological innovation. The “Porter hypothesis” states that an appropriate level of ER can incentivize enterprises to engage in Pat, enhancing productivity, offsetting environmental protection costs, and increasing profitability [35]. Porter and Linde (1995) propose that effective ER can improve firms’ environmental performance through five main pathways [36]. 1) ER can prompt firms to identify inefficiencies in resource use, leading to technological advancements that enhance resource efficiency and reduce pollution. 2) ER focused on environmental information collection can highlight high-performing companies, allowing them to gain economic benefits and motivating them to strengthen their environmental measures. 3) By clarifying relevant laws and regulations, ER can reduce investment uncertainty for firms, emphasizing the significance of energy conservation and emissions reduction and encouraging firms to enhance their environmental practices. 4) ER can exert pressure on companies, compelling them to reduce emissions through Pat. 5) ER can balance the competitive environment during the transition period. Additionally, empirical studies have confirmed that ER can promote Pat, thereby reducing CI [7, 29, 37].

The third is industrial structure. By restricting the growth of HHE and encouraging green, environmentally friendly sectors, ER optimizes the IS. On the one hand, the government enforces strict ER measures on HHE, thereby curbing pollution at its source. For example, the “Regulations on the Prevention and Control of Air Pollution in Guangdong Province” prohibit the construction and expansion of high-pollution industrial projects listed in the catalog within the Pearl River Delta region. On the other hand, in order to encourage the sustainable growth of green and low-carbon sectors and to support the green transformation of HHE, the government has implemented a variety of tax and fiscal incentives. For instance, “The Guiding Opinions of the General Office of the State Council on Accelerating the Promotion and Application of New Energy Vehicles” proposes enhancing subsidy policies for new energy vehicles and providing tax incentives for their adoption. Additionally, empirical studies have demonstrated that ER reduces CI by optimizing the IS [11, 22, 38].

The Spatial Interaction Mechanism of ER on CI

First, due to the spillover effect of CO₂ emission between different regions [39] or the inter-regional transfer of HHE [32, 40], regions implementing ER do not fully benefit from these regulations. This creates a spatial spillover effect of ER on CI. Through the spillover of CI, the carbon emission reduction effect of ER in nearby places might affect the local concentration of carbon.

Second, varying intensities of ER across regions cause production factors to migrate, driving enterprises unable to meet strict ER requirements to relocate to adjacent areas [32]. This migration impacts overall regional carbon management effectiveness. Additionally, when ER prompts the near transfer of high emissions [25], strict local ER pushes HHE to regions with weaker ER, resulting in an “extrusion effect” between regions.

Third, local governments exhibit imitative interaction behavior in ER, meaning the intensity of ER in adjacent areas influences the local region’s ER strength and consequently impacts local carbon emissions. This imitative competition behavior generally weakens the effectiveness of pollution control [25]. However, as China increasingly emphasizes environmental protection, the strategic interactions between local governments regarding ER have begun to shift towards a “race to the top” [41].

The specific spatial interaction effects can be analyzed through the three main mechanisms by which ER influences CI. First, the ES effect. Optimization of ES by local ER policies can affect the CI of other regions through “imitation effects” and “extrusion effects”. On the one hand, adjacent regions can imitate the local ER policies to reduce fossil fuel consumption [13], such as adopting centralized heating instead of decentralized heating and encouraging enterprises and institutions to use clean energy, thereby reducing regional CI. On the other hand, stringent local ER policies may push some HHEs to relocate to other regions rich in fossil fuels, increasing fossil fuel consumption and consequently raising CO₂ emissions in those areas [10].

Second, the technology effect. The improvement in technology level due to local ER policies can affect the CI in adjacent regions through “imitation effects”, “competition effects”, and “technology spillover effects”. 1) When local ER policies promote green technology advancements, other regions may imitate and implement similar regulatory measures [42]. 2) In a “Race to the top” scenario of policy interaction, local environmental regulation-driven Pat may encourage neighboring regions to enhance their research and development investments in green technology [28], thereby reducing overall regional CI levels. 3) Positive spillovers from green technology innovation can improve the air quality of adjacent areas. However, if these areas opt to reduce their own investment in green technology, their “free-riding” behavior can undermine overall regional carbon reduction efforts.

Third, the structural effect. Promoting IS upgrading by local ER policies can affect the CI of other regions through “imitation effects” and “extrusion effects”. On the one hand, neighboring regions may imitate local ER policies to achieve specific emission reduction targets, driving their own IS upgrading [43]. The clustering of high-end manufacturing and productive service industries can produce positive spillover effects, optimizing the IS of these regions. On the other hand, strict local ER policies, such as restricting the expansion of HHE industries or blacklisting certain industrial developments, may push non-compliant HHEs to relocate to other regions [37, 39], thereby increasing the CI in those areas.

In conclusion, ER can impact CI through ES, Pat, and IS. The effects are depicted in Fig. 1.

Variables

This article studies the impact of ER on CI, involving the following variables:

Fig. 1.

Mechanisms of ER effects on CI and their spatial interactions.

First, the explained variable in this study is carbon intensity (CI), which is calculated as the ratio of CO₂ emissions to GDP for each province from 2007 to 2019 [44]. Following the CO₂ emission calculation method provided by the IPCC in 2006 [45], the total yearly CO₂ emissions of 30 Chinese provinces were determined (Xizang, Hong Kong, Macau, and Taiwan were left out due to data inaccessibility). The formulas for CI are provided in Equations (1) and (2).

In Equation (1), C_pt represents the CO₂ emissions of province p in year t. E_ipt denotes the consumption of energy type i in year t in province p, σ_i represents the conversion coefficient of standard coal for the energy type i, and φ_i is the carbon emission coefficient for the energy type i. Specific coefficients are shown in Table 1. The conversion factor between carbon emissions and CO₂ is 44/12.

In Equation (2), CI_pt represents the carbon intensity of province p in year t. GDP_pt represents the GDP of province p in year t.

Second, the explanatory variable is environmental regulation (ER). Since there is no uniform standard for measuring ER [13]. Compared to other types of ER, investment-based ER has been the most effective in promoting green innovation and reducing emissions [29]. Consequently, this paper uses the completed investment in industrial pollution control to represent the level of ER [20].

Third, control variables were included following prior research [48, 49]. Energy consumption intensity (EI), population size (P), and Per capita GDP (PGDP) of each province were used as representative variables for technology, population levels, and wealth, respectively. The data for the above variables are sourced from the China Energy Statistical Yearbook, the China Environmental Statistical Yearbook, the China Statistical Yearbook, the Wind database, and the various provincial statistical yearbooks. Definitions of the variables are provided in Table 2.

Econometric Methods

(1) Modeling the carbon abatement effects of ER

This paper combines the ISIRPAT model with the DSDM to establish an empirical model. To enhance the accuracy of the research, this study constructs three models incorporating the time lag term CI_i,t–1 (Equation (3)), the spatiotemporal lag term WCI_i,t–1 (Equation (4)), and both time and spatiotemporal lag terms simultaneously (Equation (5)), as shown below:

μ represents the impact of the previous period’s CI of this province on its current period’s CI, W represents the spatial weight matrix, α represents the effect of the neighboring provinces’ previous period’s CI on this province, ρ represents the influence of the current period’s CI of the surrounding provinces on this province, and β₂, β₄, β₆, β₈ represent the effects of the neighboring provinces’ ER, PGDP, ES, and P, respectively, on the province’s CI.

(2) Modeling of the mechanism test

This study, incorporating the DSDM model, tests for the presence of mediating effects of ES effect, technological effect, and IS effect in the process by which ER influences CI through the construction of the following mediation models.

In Equation (7), M represents the mediator variables, which include the ES effect, the Pat effect, and the IS effect. The meaning of the other variables is consistent with Equations (1) to (5).

To compare the regional heterogeneity of ER, the regional division of this study is consistent with [39]. Each region is empirically tested by establishing Equations (3) to (8), respectively.

Model Testing

Table 3 presents the results of the econometric model test. All indices are significant at the 1% level, showing a high spatial connection in the study subject. The findings of the LR test show that the spatial lag model (SLM) or the spatial error model (SEM) cannot be used to simplify the spatial Durbin model (SDM). Furthermore, the estimation results demonstrate that the two-way fixed effects model fits data better than the individual fixed and temporal fixed effects models. Therefore, this paper uses a two-way fixed SDM model.

Carbon Emission Reduction Effects of ER

The SDM estimation results for the full sample are shown in Table 4. To facilitate comparison, estimations were conducted for SDM models with only a time lag, a spatiotemporal lag, and both time and spatiotemporal lags. The model incorporating both time and spatiotemporal lags, which demonstrated the highest goodness of fit, was primarily analyzed. The regression results indicate that CI exhibits significant spatial spillover, spatial lag, and time lag effects. Specifically, the CI of the previous year positively influences the current year’s CI within the province (0.7096), and the CI of neighboring provinces positively impacts the current year’s CI of the province (1.4073). This suggests that carbon emissions are continuous over time and have spillover effects across regions. However, the province’s CI in the current year can be negatively impacted by the CI of nearby provinces in the prior year (-4.0948). This may be due to stronger environmental enforcement in neighboring provinces in the previous year, reducing their CI and displacing HHE to the province in question, thereby increasing its CI in the current year. Additionally, positive environmental spillovers may occur when neighboring provinces reduce their CI in the previous year, potentially leading to a “free-riding” behavior. This could result in relaxed pollution control measures and increased CI in the current year. Most researchers’ study findings support CI’s observed time and spatial lag characteristics [50-52], and the effects of time and spatial interaction are consistent with [53].

Since the spatial lag term of the independent variable in this model does not consider the interaction information between multiple neighboring provinces. Therefore, its regression coefficients do not reflect the independent variables’ spatial spillover characteristics. Nonetheless, the direct, indirect, and total effects can effectively reveal the spatial interaction linkages between ER and related control variables on CI. Given that the DSDM model, which includes both time lag and spatiotemporal lag variables, has the highest goodness of fit, this analysis focuses on the various effects in its regression results. Table 5 presents the regression coefficients.

The short-term effects of ER on CI are significantly negative (-1.9291), (-7.2552), and (-9.1843) for direct, indirect, and total effects, respectively. This indicates that the ER within a province not only reduces its own CI but also that of neighboring provinces, as well as the overall CI across the entire sample region. This indicates that ER has a significant spatial spillover effect on decreasing carbon emissions. The main reasons are the following: First, in response to the national dual-carbon goals, provinces have intensified ER, leading to a decline in coal consumption and continuous optimization of ES. Second, both government and enterprises have increased investments in environmental pollution control and reduced provincial CI by enhancing the level of green Pat. Third, strategic interactions in ER among local governments are starting to shift towards a “race-to-the-top” scenario, where neighboring provinces emulate the environmental protection measures of other provinces and adjust economic structures. The IS effect exhibits spatial spillover, further optimizing the province’s economic structure and encouraging a decrease in CI within and throughout the sample region. Furthermore, the long-term total effect of the ER on CI is still strongly negative (-3.7809), reflecting the sustainability of the emission reduction effect through green technology innovation. These findings are consistent with previous research conclusions, highlighting the large impact of ER on carbon emissions reduction in China [4, 11, 29].

In the short term, the increase in PGDP has a favorable impact on the province’s CI (4.0563). Similarly, growth in PGDP in neighboring provinces will also increase the CI of the province (27.0010), thus having a positive total effect on the CI across the entire sample region (31.0573). This is likely because rising per capita wealth levels within a province increase resource consumption. Neighboring provinces also primarily grow their wealth through resource consumption, including resources purchased from the province, thereby elevating the CI for both the province and the sample region. Furthermore, as living standards improve with increased per capita wealth, demand for various products rises, increasing the CI of both the province and the sample region. Higher per capita wealth levels will lead to a broader range of demands in the long term, further increasing the CI across the whole sample area (12.7867).

In the short term, reducing a province’s EI dramatically lowers its CI (1.8318). Additionally, energy intensity reductions in neighboring provinces also significantly decrease the province’s CI (72.6358). Consequently, the reduction in energy intensity in the province has a substantial carbon reduction effect on the whole sample area (74.4676). This demonstrates that improvements in the province’s green technology enhance energy utilization efficiency. Furthermore, there are spatial spillovers from green technology advances in neighboring provinces, further boosting the province’s technological levels and thereby reducing its CI. Therefore, the development of green technology within the province stimulates Pat across the entire sample region, leading to a reduction in CI. In the long term, decreasing the province’s energy intensity promotes a reduction in CI for the entire sample region (30.6598), indicating that the carbon emission reduction effects driven by technological advancements are enduring.

In the short term, a population decline within the province will reduce its CI (3.3939). Similarly, a population decrease in neighboring provinces will further reduce the province’s CI (27.7662). Consequently, the province’s population drop has a considerable impact on carbon emission reduction across the entire sample region (31.1601). This implies that a fall in the province’s population will reduce consumption of local products, while a population decline in surrounding provinces reduces demand for products, including those sourced from the province. This reduction in demand leads to decreased resource consumption for production within the province, thereby lowering its CI. Furthermore, population loss has a long-term carbon-reducing effect (12.8288).

Analysis of Mechanism Test Results

Table 4 presents the consequences of ER’s dynamic spatial interaction on CI, as Equation (6) has been previously tested. This section focuses on estimating Equations (7) and (8). Suppose ER significantly affects the mediator variable, and the mediator variable significantly affects the CI. In that case, the transmission mechanism of the mediator variable is valid, and the mediation effect is effective. Otherwise, the transmission mechanism is hindered, and the mediating effect is not valid. Table 6 shows the mediation effect estimates. Other estimates are available on request.

(1) ES effect

In Equation (7), with ES as the dependent variable, the short-term optimization of ER in neighboring provinces improves the ES of the focal province by lowering the proportion of coal use in total energy consumption (-0.1689). Similarly, the ER within the focal province optimizes the ES across the entire sample area (-0.1714). This is likely because, in the short term, ER measures in neighboring provinces encourage HHE to reduce coal usage by enhancing energy efficiency through technological upgrades. The technological spillover effect further reduces coal usage both within the focal province and across the entire sample area.

In Equation (8), with ES as the independent variable, the short-term optimization of the ES within a province reduces the CI of the entire sample area (20.5372), and this emission reduction effect is sustainable (9.6603). Reducing coal consumption in the province decreases carbon dioxide emissions and has a positive spillover effect on carbon emissions; it further reduces the CI across the whole sample area.

In the short term, the analysis above indicates that ER can reduce the CI of the entire sample area by optimizing the ES. The mechanism by which the ER influences the ES is effective, and its impact on emission reduction is significant.

(2) Pat effect

In Equation (7), with LnPat as the dependent variable, the short-term effects show that ER within the province reduces the Pat (-0.2212), while ER in neighboring provinces increases the Pat in the focal province (2.6255). Additionally, the ER within the province significantly increases the Pat across the entire sample area (2.4042). The primary reason is that, in the short term, firms within the province allocate more funds to pollution control equipment and post-event management rather than RD investment, leading to a temporary decline in Pat. Meanwhile, neighboring provinces invest more in green technology RD investment, and due to the positive spillover effect, this increases the Pat in the focal province and the whole sample area. In the long term, as investments in pollution control grow, the level of Pat within the province significantly improves (0.3256), and this improvement has a lasting effect across the entire sample area (0.3307).

In Equation (8), with LnPat as the independent variable, increasing the amount of technical innovation in nearby provinces reduces the focal province’s CI in the short term (-12.5565). Due to the positive spillover effect of technological advancements, an increase in patent applications within the focal province reduces CI across the entire sample area (-12.5468), with this technological impact providing a lasting carbon reduction effect (-5.4121).

In summary, ER reduces CI by enhancing regional Pat in both the short and long term. The process by which ER fosters Pat is effective and significantly contributes to emission reduction.

(3) IS effect

In Equation (7), with LnIS as the dependent variable, the short-term effects indicate that ER in nearby provinces optimizes the focal province’s IS (-0.1197). Similarly, the ER in the province has optimized the IS of the whole sample area (-0.1203). The reason for this optimization is probably that ER in nearby provinces encourages the upgrading and transformation of HHE, which lowers their need for upstream and downstream production materials, including those from the focal province and other regions. This improves the IS of the whole sample region as well as the focal province.

In Equation (8), with LnIS as the independent variable, the short-term impact shows that reducing the value produced by the secondary industry within a province raises its CI (-8.7503). A decrease in the secondary industry’s value added in neighboring provinces reduces the province’s CI (10.9802). Additionally, a decrease in the secondary industry’s value-added within a province leads to a reduction in CI across the entire sample area (2.2298). The emission reduction effect from this decrease in secondary industry value added shows persistence (0.8677).

Therefore, throughout the entire sample region, ER can optimize the IS and reduce CI in the short run. The process by which the ER affects the IS is effective. In summary, ER optimizes ES and IS and promotes Pat in the short term. Meanwhile, ES has the best emission reduction effect, followed by Pat and IS. Only Pat can contribute to long-term pollution reductions.

Effect of ER on Reducing Carbon Intensity

Table 7 presents the DSDM model findings for subsamples, and Table 8 presents the results of dynamic interaction effects.

(1) Eastern region

As shown in Table 7, the CI exhibits significant time lag effects, spatial spillover effects, and spatial lag effects in the eastern region. These influences and theoretical analyses align with the entire sample area.

As shown in Table 8, in the short term, ER in the eastern region has a significant negative direct, indirect, and total effect on CI. The long-term total effect is also negative. This may be due to the eastern region’s more advanced IS, which allows stronger ER to drive the transformation of HHE more effectively, thereby reducing regional CI. The direct effect of PGDP on CI in the short term is not significant, indicating that the growth in per capita wealth does not impact CI. A possible reason is that increased PGDP boosts product demand. However, the higher technological level in the eastern region offsets the carbon emissions associated with this increased production. Therefore, the growth of per capita wealth does not significantly affect the province’s CI.

(2) Middle region

Table 7 indicates a notable difference for the middle region compared to the entire sample. In this region, the coefficient for the lagged CI is significantly negative (-2.6816), suggesting that within these provinces, a rise in CI in the prior year causes a decrease in CI in the current year. A possible reason is the enhanced environmental awareness in provinces of the middle region. To achieve certain emission reduction targets, efforts will be made to reduce the CI of the current year if there is an increase in the previous year.

Table 8 illustrates that the total effect is significantly negative (-1.4568); ER’s short-term direct and indirect effects are not significant in the middle region. This indicates that the ER in both the province and neighboring provinces does not impact the province’s CI in the short run, yet the ER in the province can reduce the CI across the entire sample region. The primary reason is that technological advancements within the province are slow, making it difficult to achieve significant emission reductions from industrial pollution control investments in the short term. However, due to the abundance of clean energy in the central region, substituting a portion of coal consumption with clean energy in the short term reduces the province’s coal consumption. This reduction has a favorable spillover impact on carbon emissions, thereby lowering the CI across China. In the long term, the province’s ER significantly reduces the province’s CI (-0.4235), and the province’s ER reduces the CI of the entire sample region (-0.3102). In contrast, the ER in neighboring provinces significantly increases the CI of this province (0.1133). This might be due to neighboring provinces tightening ER, prompting HHE to relocate to this nearby province, leading to an increase in the CI of this province. As the environmental awareness in this province increases, its ER effectively reduces the province’s CI and, due to the imitation effect in other provinces, effectively reduces the CI of the entire sample region.

(3) Western region

Unlike the results from the entire sample, Table 7 reveals a significant negative coefficient for the lagged CI in the western region (-0.2065). This suggests that a rise in CI in the previous year results in a decrease in CI in the current year within the western provinces. This trend suggests heightened environmental consciousness in these provinces, where efforts to meet emission reduction targets intensify after a year of increased CI, thereby driving efforts to reduce the current year’s CI.

Table 8 reveals that the direct, indirect, and total effects of ER in the short term are significantly positive, being (2.3091), (6.7940), and (11.7899), respectively. This suggests that local ER increases CI within the province, ER in neighboring provinces similarly escalates the local province’s CI, and local ER also raises the CI throughout the entire sample region. There are three possible reasons for this. First, compared to the Middle and Eastern regions, the Western region’s intensified ER prompts firms to boost their RD investments to enhance energy efficiency. The enhancement in energy efficiency paradoxically stimulates an increase in energy demand. The western region’s rich energy resources create a pull factor for industrial enterprises moving westward from coastal areas. This migration heightens resource and energy demands, subsequently elevating the CI both locally and across the entire sample region. Second, enterprises may accelerate their production activities by anticipating further stringent ER measures and rising future energy costs. This preemptive increase in production boosts fossil energy consumption in the short term, thereby amplifying the CI across both local and broader regions. Moreover, in the long term, provinces in the western region continue to introduce industrial enterprises from coastal areas for sustained economic growth. The “Green Paradox” and the “Jevons Paradox” will long-term affect the western region, thereby increasing its CI (6.6737).

In summary, ER in the eastern and middle regions of China has a significant abatement effect; on the contrary, the abatement effect of ER in the western region is not obvious.

Analysis of Mechanism Test Results

This study investigates the mediating roles of these three mechanisms in the eastern, middle, and western regions to delve deeper into how ER’s ES, technological, and IS effects manifest across various regions. Due to space constraints, Table 9 only presents the findings related to the dynamic spatial interaction effects of the mediator variables and the independent variables in the mediation model. Additional results are available upon request.

(1) Eastern region

In Equation (7), with LnIS as the dependent variable, short-term ER within the province significantly optimizes the local IS (-0.0159). Similarly, ER in neighboring provinces improves the province’s IS (-0.0668), and the province’s ER optimizes the entire eastern region’s IS (-0.0827). This improvement is likely due to the eastern region’s more sophisticated industrial base than the middle and western areas. Investments in pollution control significantly drive the transformation and upgrading of secondary industries. Equation (8) indicates that where LnIS is the independent variable, optimizing IS has a significant short-term impact on lowering carbon intensity. The optimization within the province not only significantly lowers its CI, but improvements in neighboring provinces’ IS also substantially reduce the province’s CI. Moreover, the provincial enhancements substantially decrease the CI across the eastern region. These findings indicate that the mechanisms relating ER to IS optimization in the eastern region are operating effectively in the short term.

(2) Middle region

In Equation (7), ES is the dependent variable. In the short term, ER within the province positively influenced the local ES (-0.0447), neighboring provinces’ ER optimized the province’s ES (-0.0184), and the province’s ER optimized the entire middle region’s ES (-0.2321). These short-term improvements might stem from HHE reducing coal consumption under stringent ER policies. To prevent the ER of the province from causing an “extrusion effect”, neighboring provinces might exhibit an “imitation effect” in their environmental policies. Moreover, such ER efforts also decrease the CI locally and regionally due to significant positive spillover effects. However, over the long term, strict ER promotes the advancement of green technologies, enhancing energy efficiency but increasing energy demand and thus the CI of the middle region (0.1779). In Equation (8), with ES as the independent variable, short-term optimization of the province’s ES significantly reduces the CI across the entire middle region (27.9414). Therefore, the ES effect of ER in the middle region operates smoothly in the short term.

(3) Western region

The impact of ES is demonstrated in Equation (7), where ES is the dependent variable. In the short term, ER in neighboring provinces hinders the adjustment of the province’s ES (0.2791), while the province’s own ER hampers the ES adjustment across the entire western region (0.1518). This may be due to neighboring provinces implementing strict environmental measures to meet specific emission reduction targets, resulting in HHE relocating to this province and surrounding areas. Consequently, coal consumption in both this province and the entire western region increases, leading to a rise in the province’s CI (1.3173). Over a longer period, the province’s ER continues to restrict energy structural adjustments within the province (0.0128) and across the western region (0.0114). This could be due to the province’s severe ER, which drives enterprises to increase their technology levels in order to improve energy efficiency. However, increasing energy efficiency stimulates greater energy demand, thereby increasing coal consumption in both the province and the western region.

Equation (7) demonstrates the impact of IS, with LnIS as the dependent variable. ER has hindered short-term adjustments to the IS in the province (0.0192). This limitation is because the IS in the western region is more unbalanced than in the middle and eastern regions. The province’s investment in pollution control funding was more directed towards developing the secondary industry sector, such as introducing manufacturing from coastal areas and expanding the local scale of secondary industry production. In Equation (8), with LnIS as the independent variable, the imbalance in IS within the province causes a short-term increase in CI across the entire western area (216.2943). The rising proportion of industrial output in the province elevates the demand for coal and other energy sources, resulting in positive spillover effects on carbon emissions and subsequently increasing CI throughout the western region.

In conclusion, the emission reduction mechanism of ER varies in the eastern, middle, and western regions of China. In the short term, the eastern part benefits significantly from improvements in IS, while the middle region sees notable emission reductions due to changes in ES. Conversely, ER leads to an increase in CI in the western region, driven by both energy and IS. The results align with earlier research conducted by [10, 44], which emphasizes notable decreases in emissions in the eastern and middle regions. Additionally, Zhang et al. (2021) reported a rise in CI in the western region due to ER [25].