Abstract:

[Objective] China’s carbon-peaking and carbon-neutrality targets and its shift toward coordinated pollution control and carbon mitigation have made it increasingly important to identify how industrial environmental pressure changes across development stages. Yet much of the literature on the industrial Environmental Kuznets Curve (EKC) still examines carbon emissions and conventional pollutants separately, which may obscure composite risks when pollution control and carbon constraints evolve along different trajectories. This study therefore examines whether incorporating industrial CO₂ emissions into measures of industrial environmental pressure changes the identified EKC morphology, turning-point locations, and development-stage assessments, using Hainan Province, China, as a case study. Hainan is a tropical island province characterized by strong ecological constraints, a service-led economy, and an industrial sector that, despite its modest share of aggregate output, can generate locally concentrated pressures on energy use, water resources, and waste management. [Methods] Drawing on annual industrial data for Hainan Province during 2003-2023, the study constructed a multidimensional environmental-pressure framework covering industrial water use, industrial wastewater discharge, conventional industrial air-pollutant emissions, industrial solid waste generation, and industrial CO₂ emissions. Industrial CO₂ emissions were estimated from nine industrial energy categories (raw coal, coke, crude oil, gasoline, kerosene, diesel oil, fuel oil, natural gas, and electricity) after conversion into standard coal equivalents, using the coefficient approach recommended by the Intergovernmental Panel on Climate Change. Range standardization and entropy weighting were used to generate composite indicators. The empirical design was organized around two parallel comparisons. The first contrasted conventional industrial air-pollutant emissions (CIAPE) with the Carbon-Air Pollutant Pressure Index (CAPPI), which incorporated industrial CO₂ emissions within a comparable standardized framework. The second compared the Environmental Pollution Composite Index Excluding Industrial CO₂ Emissions (EPCI−ICO₂) with the Environmental Pollution Composite Index Including Industrial CO₂ Emissions (EPCI+ICO₂). EPCI−ICO₂ combined industrial wastewater discharge, CIAPE, and industrial solid waste generation, whereas EPCI+ICO₂ replaced CIAPE with CAPPI. By applying the same polynomial model specification to both sets of indicators, the study focused the comparison on differences associated with carbon inclusion rather than on unrelated changes in model form. Linear, quadratic, and cubic polynomial EKC regressions were estimated to identify curve morphology and turning points with respect to industrial value added. Model selection considered statistical significance, adjusted R-squared, and the Akaike and Bayesian information criteria. Augmented Dickey-Fuller unit-root tests were conducted, and nonstationary series were differenced where necessary to reduce the risk of spurious regression. Standardized turning points were transformed back to actual levels of industrial value added to support development-stage interpretation. The analysis compared curve morphology, turning points, and development-stage assessments under alternative environmental-pressure boundaries rather than estimating causal effects. [Results] Hainan’s industrial environmental pressures do not conform to a single classical inverted-U-shaped pattern; instead, different environmental media evolve asynchronously. Industrial water use follows an N-shaped trajectory, indicating a potential rebound in water-resource pressure at higher levels of industrial development, whereas industrial wastewater discharge shows no stable systematic relationship with industrial value added. Industrial CO₂ emissions, CAPPI, industrial solid waste generation, EPCI−ICO₂, and EPCI+ICO₂ all exhibit inverted-N-shaped trajectories, indicating a sequence of partial easing, renewed pressure, and subsequent improvement rather than a single-peaked trajectory. Conventional industrial air-pollutant emissions are the main exception, displaying an inverted-U-shaped curve. Turning-point estimates further reveal differentiated thresholds across environmental media. Industrial CO₂ emissions and CAPPI enter a renewed improvement phase at comparatively low levels of industrial value added. By contrast, solid waste pressure enters its renewed improvement phase only at a higher level of industrial value added, while industrial water-use pressure remains exposed to rebound risk. The second turning points for solid waste generation and industrial water use are CNY 81.960 billion and CNY 84.484 billion, respectively. The CIAPE-CAPPI comparison reveals a change in both curve shape and turning-point location. Under the conventional air-pollutant specification, CIAPE follows an inverted-U-shaped curve, with its turning point located at CNY 79.293 billion of industrial value added. Under the carbon-inclusive specification, CAPPI follows an inverted-N-shaped curve, and its second turning point occurs at CNY 68.839 billion. The carbon-inclusive turning point is CNY 10.454 billion lower in terms of industrial value added, suggesting that reliance on conventional air pollutants alone may obscure multistage pressure fluctuations associated with carbon constraints. This inference is specific to the Hainan sample. At the aggregate level, both EPCI−ICO₂ and EPCI+ICO₂ follow inverted-N-shaped trajectories, indicating that Hainan’s industrial environmental pressure is better characterized by multistage fluctuation than by steady improvement after a single peak. Their second turning points occur at CNY 74.030 billion for EPCI−ICO₂ and CNY 72.516 billion for EPCI+ICO₂, a difference of CNY 1.514 billion. This difference suggests that the carbon-inclusive index identifies stage-related changes in composite environmental risk at a slightly lower level of industrial value added. Conventional air pollutants and CO₂ share some emission sources, but their abatement trajectories need not be synchronized. End-of-pipe treatment, emission standards, and local control projects can directly affect sulfur dioxide, nitrogen oxides, and particulate emissions, whereas industrial CO₂ emissions are more closely embedded in fuel structure, production processes, and the spatial configuration of industrial projects. [Conclusion] The study shifts attention from whether an inverted-U-shaped curve exists to whether alternative constructions of environmental-pressure indicators change the identified curve. Evidence from Hainan, an ecologically constrained island province, indicates that including industrial CO₂ emissions can affect not only indicator levels but also curve morphology, turning-point locations, and development-stage assessments. The analysis identifies measurement-related differences within the Hainan sample; it neither establishes a universal EKC law nor estimates causal effects. For environmental governance, the results suggest that Hainan and comparable island regions should avoid assessing industrial environmental pressure solely through conventional pollutant indicators. Carbon and air-pollution pressures should be considered jointly in project approval, environmental assessment, and dynamic monitoring of key industrial parks. Water-use rebound risk and the relatively high turning point for solid waste pressure also call for tailored policy responses; these findings caution against assuming that all environmental media improve simultaneously with industrial upgrading. More broadly, environmental-risk identification should be integrated earlier into industrial development strategies, project screening, and spatial planning. In ecologically constrained regions with service-led economies where industry retains an important supporting role, excluding industrial CO₂ emissions may underestimate composite environmental risks and misidentify the stage of green transformation.

Key words: Environmental Kuznets Curve (EKC), industrial CO₂ emissions, environmental pollution composite index, turning points, Hainan Province

摘要：

【目的】现有环境库兹涅茨曲线研究多将碳排放与常规污染物分开考察，可能影响复合环境压力识别。该文检验纳入工业二氧化碳排放是否影响曲线形态、拐点识别和阶段判断。【方法】基于2003—2023年海南省年度工业数据，构建涵盖工业用水、废水、传统工业废气、固体废物和二氧化碳排放的指标体系，设置传统工业废气与碳-气复合压力、不含碳与含碳环境污染综合指数两组比较口径，在采用熵值法构建综合污染指数的基础上，利用多项式回归模型识别曲线形态及拐点。【结果】海南各类工业环境压力并非均为倒U形，而呈现多介质异步演进特征。传统工业废气为倒U形，而碳-气复合压力转为倒N形；后者第二拐点约688亿元，较传统工业废气约793亿元的拐点低约105亿元，表明含碳口径在较低工业增加值水平识别出复合压力转折。两类综合指数均为倒N形，但含碳指数的第二拐点更低，提示其对风险识别更敏感。碳排放与碳-气复合压力的再改善拐点较低，固体废物拐点较高，工业用水存在高发展阶段反弹风险。【结论】纳入工业二氧化碳排放可影响环境压力表征、曲线识别和阶段判断。对海南及类似海岛型地区，忽略碳排放可能低估复合环境风险并误判绿色转型阶段。

关键词: 环境库兹涅茨曲线（EKC）, 工业碳排放, 环境污染综合指数, 拐点, 海南省