Model Construction and Scenario Analysis for Carbon Dioxide Emissions from Energy Consumption in Jiangsu Province: Based on the STIRPAT Extended Model

Abstract

Against the backdrop of China’s “dual carbon” strategy (carbon peaking and carbon neutrality), provincial-level carbon emission research is crucial for the implementation of related policies. However, existing studies insufficiently cover the driving mechanisms and scenario prediction for energy-importing provinces. This study can provide theoretical references for similar provinces in China to conduct research on carbon dioxide emissions from energy consumption. The carbon dioxide emissions from energy consumption in Jiangsu Province between 2000 and 2023 were calculated using the carbon emission coefficient method. The Tapio decoupling index model was adopted to evaluate the decoupling relationship between economic growth and carbon dioxide emissions from energy consumption in Jiangsu. An extended STIRPAT model was established to predict carbon dioxide emissions from energy consumption in Jiangsu, and this model was applied to analyze the emissions under three scenarios (baseline scenario, low-carbon scenario, and enhanced low-carbon scenario) during 2024–2030. The results show the following: (1) During 2000–2023, the carbon dioxide emissions from energy consumption in Jiangsu Province ranged from 215.22428 million tons to 783.94270 million tons, with an average of 549.96280 million tons. (2) The decoupling status between carbon dioxide emissions from energy consumption and economic development in Jiangsu was dominated by weak decoupling, accounting for 91.304%, while a small proportion (8.696%) of expansive coupling was also observed. (3) Under the baseline scenario, the carbon dioxide emissions from energy consumption in Jiangsu in 2030 will reach 796.828 million tons; under the low-carbon scenario, the emissions will be 786.355 million tons; and under the enhanced low-carbon scenario, the emissions will be 772.293 million tons. Furthermore, countermeasures and suggestions for reducing carbon dioxide emissions from energy consumption in Jiangsu are proposed, mainly including strengthening the guidance of policies and institutional systems, optimizing the energy consumption structure, intensifying technological innovation efforts, and enhancing government promotion and publicity.

1. Introduction

Large-scale greenhouse gas emissions are the primary driver of global warming, and fossil energy consumption constitutes the main source of carbon dioxide emissions [1,2,3]. At present, global warming and environmental deterioration have become key issues of concern to policymakers worldwide, who are committed to achieving environmental governance by reducing carbon dioxide emissions [4,5]. Against this backdrop, there is an urgent need in current research to construct carbon dioxide emission prediction models and further improve the global ecosystem by integrating the optimization of carbon peaking paths, industrial technology upgrading, and the implementation of policy supervision [6,7,8].

Scholars at home and abroad have successively adopted the IPAT (impact, population, affluence, and technology) model, IPBAT (impact, population, affluence, behavior, and technology) model, ImPACT (the emphasis on the expression Impact in the IPAT equation) model, and STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) model to predict carbon dioxide emissions under different scenarios in countries, provinces, or key cities [9,10,11,12,13,14]. Among these models, the IPAT model clearly defines a linear identity between environmental pressure and population, affluence, and technology. However, it has the limitation of being unable to quantify the elasticity coefficients of various factors [15]. The IPBAT model was constructed by incorporating the “behavioral factor (B, Behavior)” into the IPAT identity [16]. The ImPACT model focuses on decomposing “technology” into “energy consumption per unit of GDP” and “carbon emissions per unit of energy”. Meanwhile, a limitation of both the IPAT and ImPACT models is that they cannot be applied to the nonlinear relationships of variables [17,18]. The STIRPAT model profoundly explains the core breakthrough of converting the IPAT model into a logarithmic form: it quantifies the impact intensity of various factors through elasticity coefficients and allows for nonlinear relationships of variables [19]. Furthermore, it supplements the expansion of the model (e.g., incorporating variables such as urbanization rate and industrial structure) and has been widely applied in fields such as case studies and environmental governance [20,21,22].

In existing studies, Shi et al. conducted an analysis of the relationship between demographic variables and CO₂ emissions across multiple countries over a 20-year period based on the STIRPAT framework, confirming that population dynamics constitute a core driving factor behind the trajectory of carbon emissions [23]. Shahbaz et al. focused on energy-related CO₂ emissions in Malaysia and explored the impacts of urbanization, affluence, and trade openness, revealing that urbanization is the dominant inducement for emission growth [24]. Abou-Ali et al. targeted the Arab region, incorporating “fertility rate, industrialization level, urbanization level, government effectiveness, and energy production and consumption” into an extended STIRPAT model. Their findings indicated that in addition to regulating energy production and consumption, enhancing government effectiveness is also crucial for emission reduction [25]. Roy et al. integrated “energy intensity, energy demand, and energy structure” into the STIRPAT model and applied it to a study on India. The results showed that all driving factors in the model have a significant impact on environmental degradation, among which economic growth contributes the most [26]. Wang and He applied this framework to analyze the determinants of energy consumption in 30 provinces in mainland China, pointing out that population exerts a positive linear impact on energy use, while its marginal effects exhibit significant heterogeneity [24,27]. Lu et al. further integrated the STIRPAT model with the Tapio decoupling index [17], and their study on the Suzhou–Wuxi–Changzhou metropolitan area showed that economic expansion has the most prominent impact on energy-induced carbon footprints [28]. Zhao et al. combined the STIRPAT model with scenario analysis to predict the carbon emission peaks of Henan Province, covering four aspects, including energy consumption, food consumption, agricultural activities, and waste disposal [29]. Wang et al., based on the extended STIRPAT model, showed that Jilin Province’s carbon emission peak may occur between 2029 and 2045 [30]. Zheng et al. found that the growth of per capita GDP had the strongest promoting effect on CO₂ emissions, followed by energy structure and the number of civil vehicles, while energy intensity was the key factor inhibiting the growth of emissions [31]. Liu et al. screened the driving factors of carbon emission reduction in coastal cities through the extended STIRPAT model, pointing out that per capita GDP, energy structure, energy intensity, urbanization rate, and industrial structure are the core influencing factors [32]. Dan et al., based on China’s economic and social development and energy consumption data from 1980 to 2022, combined the extensible STIRPAT model with ridge regression analysis to identify seven main factors affecting CO₂ emissions from energy consumption [33]. Among them, five factors—population, urbanization rate, the proportion of the secondary industry, per capita GDP, and electrification rate—exerted positive driving effects, while two factors—the proportion of fossil energy and energy intensity—had negative driving effects [33]. Huang et al. identified key factors through ridge regression and constructed a STIRPAT carbon emission prediction model, confirming that energy structure, industrial structure, and urbanization rate are the significant factors affecting carbon emissions in Guangxi [34]. Lian et al. (2025), based on the improved STIRPAT model, found that population, per capita GDP, and industrial structure had positive driving effects on carbon emissions in Fujian Province, while energy intensity, energy structure, and foreign trade degree showed negative driving effects [35]. Existing carbon emission studies mostly focus on the national or regional level, and there is insufficient analysis of the driving mechanisms of carbon emissions in economically developed and energy-importing provinces. In addition, provincial-scale studies mostly concentrate on 3–5 driving factors, with insufficient exploration of the interaction effects between energy consumption structure and technological innovation [36]. Moreover, the application of ridge regression in the multi-factor extended STIRPAT model remains relatively scarce [37].

The Action Plan for Carbon Peaking Before 2030, issued by the Chinese government, points out that provincial-level units serve as key carriers for the implementation of carbon reduction policies, and research on the driving mechanisms and prediction of their carbon emissions plays a supporting role in the achievement of national goals. Jiangsu Province has ranked second in China in terms of GDP for 14 consecutive years, yet its per capita carbon emissions are higher than the national average. Although Jiangsu has the fiscal and technological foundations required for early carbon peaking, its rapid economic expansion has simultaneously driven up energy demand and CO₂ emissions [38,39]. Empirical studies have shown that population aging, economic growth, and urbanization are all statistically significant driving factors of provincial CO₂ emissions [40], and the impact intensity of these factors varies significantly depending on the stage of socioeconomic development and urbanization paths [41,42]. As an “energy-importing province”, Jiangsu has long had a coal consumption ratio exceeding 60%, while the installed capacity of renewable energy has grown rapidly, showing typical characteristics of energy structure transformation [43]. Meanwhile, Jiangsu is China’s first “pilot province for carbon peaking” and has issued the Implementation Plan for Carbon Peaking in Jiangsu Province.

Consequently, this study will conduct the following work: (i) The carbon emission factor method is adopted to clarify the carbon emissions generated from energy consumption in Jiangsu Province from 2000 to 2023. (ii) The Tapio Decoupling Index Model is applied to identify the decoupling state between economic development and carbon emissions from energy consumption in Jiangsu Province. (iii) Based on the extended STIRPAT model, a prediction model for CO₂ emissions from energy consumption in Jiangsu Province is constructed by incorporating factors such as urbanisation rate, proportion of tertiary industry, population size, carbon emission intensity, energy structure, and per capita GDP. (iv) Under three scenarios (baseline scenario, low-carbon scenario, and enhanced low-carbon scenario), the variation trends in CO₂ emissions from energy consumption in Jiangsu Province from 2024 to 2030 are predicted. (v) Targeted emission reduction strategies are proposed in combination with policy analysis to provide scientific support for Jiangsu Province to achieve carbon peaking. This study aims to offer guidance for the trends in CO₂ emissions from energy consumption in Jiangsu Province, and meanwhile, provide a theoretical reference for similar provinces that conduct comparable CO₂ emission research.

2. Research Methods and Data Sources

2.1. Research Methods

2.1.1. Calculation of CO₂ Emissions from Energy Consumption

The most widely adopted approach for estimating CO₂ emissions is the carbon-emission-factor method [10]. Its underlying principle is to quantify the carbon released during the combustion of each energy carrier on the basis of its intrinsic physicochemical properties [2,6]. By integrating differentiated emission factors with the corresponding energy-consumption data, this method yields precise estimates of CO₂ emissions [11]. Notably, it is also the recommended procedure in the IPCC Guidelines for National Greenhouse Gas Inventories (see Equation (1)).

$${\mathit{TE}}_{{\mathrm{C}\mathrm{O}}_2}=\sum _{\mathit{i}=1}^{\mathit{n}}\beta {\times \mathit{FEC}}_{\mathit{i}}\times {\mathit{EF}}_{\mathit{i}}$$

(1)

where ${\mathit{TE}}_{{\mathrm{C}\mathrm{O}}_2}$ denotes total CO₂ emissions from energy consumption (10⁴ t); FEC_i represents the consumption of the i energy type (10⁴ t); EF_i is the standard-coal conversion factor for the i energy type (kg ce kg⁻¹); β is the CO₂ emission coefficient per tonne of standard coal (2.493 t CO₂ t ce⁻¹); n is the number of energy categories; and i indexes individual fuels. The standard-coal conversion factors adopted for the principal fuels are: Coal (0.714), coke (0.971), crude oil (1.429), gasoline (1.471), kerosene (1.471), diesel (1.457), fuel oil (1.429) and liquefied petroleum gas (1.714). For example, the calculation process of CO₂ emissions from energy consumption in Jiangsu Province in 2023 is as follows: ${\mathit{TE}}_{{\mathrm{C}\mathrm{O}}_2}$ = β (2.493)*FEC_coal (8243.970)*EF_coal (0.714) + β (2.493)*FEC_coke (311.110)*EF_coke (0.971) + β (2.493)*FEC_{crude oil} (1383.590)*EF_{crude oil} (1.429) + β (2.493)*FEC_gasoline (26.940)*EF_gasoline (1.471) + β (2.493)*FEC_kerosene (26.010)*EF_kerosene (1.471) + β (2.493)*FEC_diesel (62.900)*EF_diesel (1.457) + β (2.493)*FEC_{fuel oil} (165.030)*EF_{fuel oil} (1.429) + β (2.493)*FEC_{liquefied petroleum gas} (35.200)*EF_{liquefied petroleum gas} (1.714).

2.1.2. Scope and Definition of Decoupling

In 2005, Tapio introduced the Tapio decoupling-index model while investigating the elasticity among economic growth, transport activity, and CO₂ emissions in Europe [44], and, for the first time, delineated eight distinct decoupling states (

Table 1. Identification table for decoupling status.

△TE	△GDP	DI	Decoupling Status
<0	>0	DI < 0	Strong decoupling
>0	>0	0 ≤ DI < 0.800	Weak decoupling
<0	<0	DI ≥ 1.200	Recessive decoupling
>0	<0	DI < 0	Strong negative decoupling
<0	<0	0 ≤ DI < 0.800	Weak negative decoupling
>0	>0	DI ≥ 1.200	Expansive negative decoupling
>0	>0	0.800 ≤ DI < 1.200	Expansive coupling
<0	<0	0.800 ≤ DI < 1.200	Recessive coupling

). This parsimonious taxonomy has since provided scholars with a lucid and unambiguous framework for characterizing the nexus between economic development and environmental protection. Second, the Tapio framework accommodates multi-dimensional drivers, yielding a more comprehensive representation and, consequently, a more accurate characterization of the decoupling between economic growth and carbon emissions [45,46]. Scholars predominantly employ the Tapio decoupling index model by constructing an elasticity indicator to scrutinize the decoupling relationship among variables [7,44,45,46]. Notably, the model is dimension-free; its estimates remain invariant to changes in measurement units, thereby preserving both applicability and cross-context comparability and enhancing its operational robustness and generalizability. The precise formulation is given in Equation (2).

$$\mathit{DI}={\displaystyle \frac{\Delta \mathit{TE}/\mathit{TE}}{\Delta \mathit{GDP}/\mathit{GDP}}}={\displaystyle \frac{({\mathit{TE}}_1-{\mathit{TE}}_0)/{\mathit{TE}}_0}{({\mathit{GDP}}_1-{\mathit{GDP}}_0)/{\mathit{GDP}}_0}}$$

(2)

where DI denotes the decoupling index, ΔTE/TE is the percentage change in energy-related CO₂ emissions between two periods, and ΔGDP/GDP is the corresponding percentage change in economic output. Specifically, TE₀ and TE₁ represent CO₂ emissions at times t₀ and t₁, respectively, while GDP₀ and GDP₁ denote gross domestic product at the same two time points.

The core value of the Tapio decoupling-index model lies in its analytical logic characterized by “quantification, dynamics, and multidimensionality,” which breaks the limitations of the linear cognition of the “growth-environmental pressure” relationship [7,10]. Against the backdrop of the “dual carbon” goals, the decoupling model has become a key bridge connecting “macro development goals” and “micro practical actions,” and plays an irreplaceable role in promoting the transformation of the economy and society toward “green, low-carbon, and sustainable” development [45,46].

2.1.3. STRIPAT Extended Model

The IPAT model is commonly used to analyze issues related to carbon emissions [11,12]. First proposed by Ehrlich and Holdren, the model attributes environmental pressure to the product of three factors: population, income, and technology [15]. The canonical specification is presented in Model (3):

$$\mathit{I} = \mathit{PAT}$$

(3)

In Equation (3), I, P, A, and T denote environmental impact, population, affluence, and technology level, respectively;

York, Rosa, and Dietz assigned different exponents to population size, per capita income, and technology level, and on this basis, proposed the STIRPAT model [18,19]. The canonical specification is presented in Model (4):

$$I = a{P}^b{A}^c{T}^{d}e$$

(4)

In Equation (4), a is the constant coefficient; b, c, and d are the elasticities to be estimated; and e represents the stochastic error term.

Taking the natural logarithm of both sides of Model (4) yields Model (5):

$$\ln I = \ln a + b\ln P + c\ln A + d\ln T + \ln e$$

(5)

To identify the determinants of Jiangsu’s energy-related CO₂ emissions (TE), the standard STIRPAT model is extended to reflect the province’s specific socio-economic context. The population dimension (P) is decomposed into demographic structure and scale, represented by urbanization rate (UR) and year-end population size (PS). The affluence dimension (A) is split into development level and economic structure, proxied by per capita GDP (PCG) and the tertiary-industry share of output (PTI). The technology dimension (T) is operationalized through advances in new-energy deployment and improvements in fossil-energy conversion efficiency, captured by energy structure (ES) and carbon intensity (CI). Extending Model (5) accordingly yields Model (6):

$$\ln \mathit{TE} = \ln \mathit{a}+{\alpha }_1\ln \mathit{UR}+{\alpha }_2\ln \mathit{PS}+{\alpha }_3\ln \mathit{PCG}+{\alpha }_4\ln \mathit{PTI}+{\alpha }_5\ln \mathit{ES}+{\alpha }_6\ln \mathit{CI}+\ln \mathit{e}$$

(6)

In the equation, α₁, α₂, α₃, α₄, α₅ and α₆ denote elasticity coefficients, indicating that a 1% change in UR, PS, PCG, PTI, ES and CI induces a α₁%, α₂%, α₃%, α₄%, α₅% and α₆% change, respectively.

2.2. Data Sources

The data is sourced from the Jiangsu Statistical Yearbook from 2001 to 2024, which calculates parameters such as urbanization rate (UR), population size (PS), industry proportion, and gross domestic product (GDP) that can represent the activity level, energy consumption, and economic development of Jiangsu Province. The data on gross domestic product (GDP) is extracted from the annual statistical yearbook of Jiangsu Province, with a value of CNY 100 million. The carbon emission intensity (CI) uses the ratio of CO₂ emissions from energy consumption to GDP in Jiangsu Province, measured in tons per CNY 10,000. The population size (PS) is based on the annual resident population of Jiangsu Province, ten thousand people. The energy structure (ES) adopts the ratio of fossil energy consumption to total energy consumption, %. Per capita GDP (PCG) adopts the ratio of Jiangsu Province’s GDP to its population size, measured in CNY per person. The urbanization rate (UR) is calculated as the ratio of non-agricultural household registration to the total population, expressed as a percentage. The proportion of the tertiary industry (PTI) refers to the ratio of the gross domestic product (GDP) of the tertiary industry region to the gross domestic product (GDP), expressed as a percentage.

For the purpose of this study, energy-related CO₂ emissions are computed exclusively from the combustion of fossil fuels. any CO₂-equivalent releases attributable to electricity consumption or to clean-energy sources (wind, hydro, solar, and nuclear) are deliberately excluded.

3. Results and Discussion

3.1. CO₂ Emissions from Energy Consumption

Energy-related CO₂ emissions in Jiangsu Province rose from 215.224 million tons in 2000 to 783.943 million tons in 2023, averaging 549.963 million tons (Figure 1A). Annual growth rates ranged from −3.850% to 23.682%, with a mean of 6.007%. In terms of fuel-specific contributions (Figure 1B), coal dominated throughout the period, accounting for 60.229–71.099% (mean 66.686%). Coke, crude oil, and fuel oil contributed 3.501–17.463% (11.295%), 16.619–24.303% (19.602%), and 0.119–2.946% (1.079%), respectively. Gasoline, kerosene, diesel, and liquefied petroleum gas (LPG) together represented minor shares: gasoline 0.046–0.475% (0.246%), kerosene 0.002–1.016% (0.079%), diesel 0.237–1.337% (0.652%), and LPG 0.216–0.709% (0.360%).

3.2. Decoupling Status of CO₂ Emissions from Energy Consumption

To neutralize scale and unit effects, all explanatory variables and the CO₂ emission series were subjected to natural-logarithmic transformation prior to estimation (

Table 2. Logarithmic observations of variables used in the STIRPAT model.

Year	lnPS	lnUR	lnCI	lnPCG	lnPSI	lnES	lnTE
2000	8.899	3.726	0.923	9.365	3.581	4.585	9.977
2001	8.904	3.752	0.872	9.461	3.597	4.575	10.027
2002	8.910	3.800	0.865	9.570	3.603	4.592	10.134
2003	8.917	3.846	0.864	9.722	3.586	4.577	10.293
2004	8.926	3.875	0.901	9.889	3.558	4.581	10.505
2005	8.934	3.922	0.800	10.081	3.581	4.546	10.605
2006	8.943	3.949	0.768	10.231	3.600	4.585	10.731
2007	8.952	3.974	0.678	10.424	3.627	4.586	10.844
2008	8.957	3.995	0.468	10.593	3.651	4.490	10.808
2009	8.963	4.018	0.408	10.695	3.679	4.474	10.856
2010	8.971	4.104	0.338	10.870	3.723	4.503	10.968
2011	8.990	4.127	0.246	11.017	3.747	4.509	11.042
2012	9.002	4.143	0.158	11.099	3.770	4.471	11.049
2013	9.011	4.165	0.119	11.191	3.809	4.520	11.111
2014	9.022	4.185	0.041	11.268	3.839	4.508	11.121
2015	9.026	4.212	−0.065	11.359	3.863	4.480	11.109
2016	9.034	4.233	−0.093	11.433	3.902	4.507	11.164
2017	9.039	4.251	−0.227	11.532	3.906	4.465	11.134
2018	9.041	4.265	−0.301	11.611	3.920	4.472	11.142
2019	9.044	4.284	−0.359	11.666	3.942	4.442	11.140
2020	9.045	4.296	−0.440	11.706	3.955	4.398	11.101
2021	9.048	4.303	−0.509	11.826	3.934	4.393	11.155
2022	9.050	4.309	−0.557	11.873	3.932	4.361	11.156
2023	9.051	4.317	−0.492	11.921	3.944	4.376	11.270

).

According to Equation (2), calculate the CO₂ emissions generated by energy consumption in Jiangsu Province from 2001 to 2023 and the decoupling elasticity index of economic development, and divide the decoupling status (calculate the actual GDP of each year based on the year 2000). Among them, weak decoupling accounted for 21 stages (91.304%), and expansive coupling accounted for 2 stages (8.696%) (

Table 3. Decoupling status between CO₂ emissions from energy consumption and economic development in Jiangsu province from 2001 to 2023.

Year	TE/(×104 tons)	GDP/(×102 million yuan, CNY)	△TE/%	△GDP/%	DI	Decoupling States
2000	21,522.428	8553.690	-	-	-	-
2001	22,622.851	9456.840	5.113	10.559	0.484	Weak decoupling
2002	25,187.477	10,606.850	17.029	24.003	0.709	Weak decoupling
2003	29,519.724	12,442.870	37.158	45.468	0.817	Expansive coupling
2004	36,510.528	14,823.130	69.639	73.295	0.950	Expansive coupling
2005	40,331.369	18,121.330	87.392	111.854	0.781	Weak decoupling
2006	45,772.292	21,240.790	112.673	148.323	0.760	Weak decoupling
2007	51,214.816	25,988.360	137.960	203.826	0.677	Weak decoupling
2008	49,400.793	30,945.450	129.532	261.779	0.495	Weak decoupling
2009	51,846.231	34,471.670	140.894	303.003	0.465	Weak decoupling
2010	58,003.913	41,383.870	169.504	383.813	0.442	Weak decoupling
2011	62,472.534	48,839.210	190.267	470.972	0.404	Weak decoupling
2012	62,901.700	53,701.920	192.261	527.822	0.364	Weak decoupling
2013	66,873.566	59,349.410	210.716	593.846	0.355	Weak decoupling
2014	67,571.319	64,830.510	213.958	657.924	0.325	Weak decoupling
2015	66,787.552	71,255.930	210.316	733.043	0.287	Weak decoupling
2016	70,510.040	77,350.850	227.612	804.298	0.283	Weak decoupling
2017	68,457.434	85,869.760	218.075	903.891	0.241	Weak decoupling
2018	69,000.830	93,207.550	220.600	989.677	0.223	Weak decoupling
2019	68,881.762	98,656.820	220.046	1053.383	0.209	Weak decoupling
2020	66,230.120	10,2807.680	207.726	1101.910	0.189	Weak decoupling
2021	69,926.452	11,6364.200	224.900	1260.398	0.178	Weak decoupling
2022	69,970.723	12,2089.280	225.106	1327.329	0.170	Weak decoupling
2023	78,394.270	12,8222.160	264.245	1399.027	0.189	Weak decoupling

). These trajectories are closely linked to the suite of policy measures implemented since China’s 10th Five-Year Plan, under which Jiangsu has vigorously restructured its energy portfolio, promoted circular-economy initiatives, strengthened environmental governance, and tightened regulatory oversight. Nevertheless, the determinants of energy-related CO₂ emissions exhibit pronounced temporal and spatial heterogeneity [47]. Moreover, as Jiangsu’s economy is still undergoing industrial transformation, a residual cohort of high-energy, high-polluting enterprises persists, rendering the decoupling elasticity inherently unstable [48,49].

3.3. Prediction Model for CO₂ Emissions Generated by Energy Consumption

Pearson correlation analysis between provincial CO₂ emissions (TE) and the six candidate drivers yields the following coefficients: r_TE–PS = 0.920, r_TE–UR = 0.953, r_TE–CI = –0.857, r_TE–PCG = 0.954, r_TE–PTI = 0.835, and r_TE–ES = −0.748, all of which are statistically significant. However, to ascertain whether these correlations can underpin a reliable forecasting model, multicollinearity diagnostics were performed. Variance-inflation factors (VIF) are 259.701 for UR, 151.739 for PS, 265.508 for PCG, 140.165 for PTI, and 27.253 for ES—far exceeding the critical threshold of 10—indicating severe multicollinearity. Consequently, the direct use of simple correlation-based specifications for projecting Jiangsu’s energy-related CO₂ emissions is inappropriate.

This paper compares the advantages and disadvantages of ridge regression, LASSO (Least Absolute Shrinkage and Selection Operator) regression, and the grey prediction model in collinearity handling (

Table 4. Characteristics of common collinearity handling methods.

Method	Collinearity Handling Capability	Data Requirements	Prediction Accuracy	Applicable Scenarios
Ridge regression	Strong (by compressing coefficients via the λ value)	Large sample (≥20 observations)	Relatively high (suitable for linear relationships)	Multivariate collinearity and linear prediction
LASSO regression	Relatively strong (by compressing coefficients to zero)	Large sample	Relatively high (suitable for variable selection)	Scenarios requiring the elimination of redundant variables
grey forecasting model	Weak (incapable of handling collinearity)	Small sample (≥4 observations)	Relatively low (suitable for trend extrapolation)	Scarce data and absence of obvious driving factors

).

Among these models, the core advantage of LASSO regression lies in variable selection. However, the six extended variables in this study all have explicit theoretical support, and forced selection may lead to the elimination of key variables [50]. The grey prediction model relies on data trend extrapolation and fails to consider the driving relationships between variables (e.g., the impact of per capita GDP growth on carbon emissions). Furthermore, this study uses 24 years of continuous data (2000–2023), which constitutes a large sample and thus does not align with the “small sample” applicable scenario of grey prediction [51]. In addition, the collinearity level in this study is “moderate”, and the “moderate coefficient shrinkage” of ridge regression is sufficient to address this issue, making radical variable elimination unnecessary [10]. The ridge trace (Figure 2) indicates that the estimated coefficients stabilize when the biasing parameter reaches k = 0.136. Consequently, the final ridge estimates reported in Table 4 are obtained at k = 0.136.

Ridge-regression results reveal that lnPS, lnUR, and lnPCG are significant at the 1% level, while lnPTI is significant at the 5% level. The model attains an R² of 0.926, and the F-statistic is significant at the 1% level (

Table 5. Results of ridge regression analysis.

Factors	Unstandardized Coefficient	Standard Error	t-Statistic	p-Value	VIF
lna	−11.811	4.315	−2.737	0.014 **	-
lnPS	1.848	0.267	6.911	0.000 ***	0.254
lnUR	0.849	0.097	8.732	0.000 ***	0.416
lnCI	0.025	0.027	0.931	0.365	0.033
lnPCG	0.202	0.022	9.388	0.000 ***	0.428
lnPTI	−0.309	0.168	−1.836	0.084 *	0.116
lnES	0.345	0.439	0.785	0.443	0.062

Note: R² = 0.926, F-Statistic = 35.432, Sig.(F) = 0.000, *** represents p < 0.01, ** represents p < 0.05, * represents p < 0.10.

). These findings confirm the systematic relationship between Jiangsu’s energy-related CO₂ emissions and the selected drivers, yielding the predictive specification presented in Model (7):

$$\ln \mathit{TE} = 1.848\ln \mathit{PS} + 0.849\ln \mathit{UR} + 0.025\ln \mathit{CI} + 0.202\ln \mathit{PCG} - 0.309\ln \mathit{PTI} + 0.345\ln \mathit{ES} - 11.811$$

(7)

Table 5 indicates that population size exerts the largest influence: a 1% increase in PS raises Jiangsu’s energy-related CO₂ emissions by 1.848%. Elasticities for the remaining drivers are 0.849% (UR), 0.025% (CI), 0.202% (PCG) and 0.345% (ES). By contrast, a 1% expansion in the tertiary-industry share (PTI) reduces emissions by 0.309%.

To validate the predictive power of Model (6), we generated retrospective forecasts for 2000–2023 and subjected them to an independent-samples t-test against observed values. At the 95% confidence level, the test yields p = 0.874, and the correlation between actual and predicted emissions reaches 0.998 (

Table 6. Independent sample t-test of model-predicted values and actual values.

Classification	The Average Value of CO2 Emissions/104 t	p-Value	t-Statistic	Sig.
Predicted value	54,609.224	0.874	−0.009	0.993
Actual value	54,996.280		−0.009	0.993

). Consequently, the model is deemed unbiased and highly accurate, and is therefore employed to project Jiangsu’s energy-related CO₂ emissions for 2024–2030.

3.4. Scenario Parameter Settings in Predictive Models

Scenario design is the cornerstone of any robust CO₂ emission projection. Drawing exclusively on authoritative provincial documents—the 14th Five-Year Plan and the 2035 Long-Range Objectives for Jiangsu Province, ancillary policy briefs, and the Jiangsu Population Forecast Compendium (2011–2030)—we derived internally consistent trajectories for all six exogenous drivers (

Table 7. The growth rate settings of various influencing factors in the CO₂ emission prediction model for energy consumption in Jiangsu Province under different scenarios (%/year).

Year	Scenario	PS	UR	CI	PCG	PTI	ES
2024–2025	BS	0.360	1.400	−2.500	6.000	4.500	−1.000
	LCS	0.320	1.200	−3.500	5.500	4.500	−1.200
	SLCS	0.280	1.000	−4.000	5.000	4.000	−1.500
2026–2030	BS	0.320	1.100	−3.000	5.000	5.000	−0.600
	LCS	0.280	1.000	−4.500	4.500	4.500	−0.800
	SLCS	0.240	0.900	−5.000	4.000	4.500	−1.000

).

Urbanisation rate (UR). Pursuant to the Outline of the 14th Five-Year Plan and the 2035 Long-Range Objectives for Jiangsu Province, the province targets 75% urbanisation by 2025 and over 80% by 2030. Accordingly, the baseline scenario assumes annual increments of 1.400% during 2024–2025 and 1.100% during 2026–2030. These trajectories are closely aligned with historical trends; proportional downward adjustments are applied for the low-carbon and enhanced low-carbon scenarios.

Proportion of tertiary industry (PTI). According to the Action Plan for Accelerating the Building of a Manufacturing-Strong Province issued by Jiangsu, the combined value added of manufacturing and producer services is expected to reach approximately 70% of regional GDP. Under the baseline scenario, the tertiary sector’s share is therefore assumed to rise by 4.500% per annum in 2024–2025 and by 5.500% per annum in 2026–2030. Correspondingly, the low-carbon and enhanced low-carbon scenarios apply additional increments of +0.500 and +1.000 percentage points per year, respectively.

Population size (PS). The Outline of the 14th Five-Year Plan and the 2035 Long-Range Objectives for Jiangsu Province set a permanent-resident target of 90 million by 2035. Cross-referencing the Jiangsu Population Forecast Compendium (2011–2030) and accounting for net migration, the province’s total population is projected to reach 88.940 million by 2030. Consequently, the baseline scenario assumes annual growth rates of 0.360% in 2024–2025 and 0.320% in 2026–2030; the low-carbon and enhanced low-carbon scenarios apply symmetric reductions of 0.400 and 0.800 percentage points per annum, respectively.

The carbon emission intensity (CI). Aligned with the national mandate of cutting carbon intensity by at least 65% relative to 2005 by 2030—the target reaffirmed in both the 14th Five-Year Plan of the People’s Republic of China and the corresponding Jiangsu provincial plan—the baseline scenario assumes annual declines of 2.500% for 2024–2025 and 3.000% for 2026–2030. For the low-carbon and enhanced low-carbon scenarios, these rates are accelerated by 1.000 and 1.500 percentage points per annum, respectively.

The energy structure (ES). In accordance with the Jiangsu Provincial Government’s Notice on Several Policy Measures for Accelerating the Comprehensive Green Transformation of Economic and Social Development, the province targets approximately 100 GW of renewable-power capacity and a non-fossil share of around 25% in total energy consumption by 2030. Under the baseline scenario, the fossil-energy share is therefore set to decline by 1.000% per annum in 2024–2025 and by 0.600% per annum in 2026–2030. Correspondingly, the low-carbon and enhanced low-carbon scenarios apply steeper annual reductions of 0.300% and 0.500%

Per capita GDP (PCG). The Outline of the 14th Five-Year Plan and the 2035 Long-Range Objectives for Jiangsu Province target a doubling of real per capita GDP relative to the 2020 level (CNY 127,000). This implies a 2035 benchmark of CNY 254,000 per capita. Under the baseline scenario, annual growth rates are therefore set at 6.000% for 2024–2025 and 5.000% for 2026–2030; the low-carbon and enhanced low-carbon scenarios apply symmetric reductions of 0.500 and 1.000 percentage points per annum, respectively.

3.5. Characteristics of Scenario Analysis

Based on the extended STIRPAT model, CO₂ emissions from energy consumption in Jiangsu Province from 2024 to 2030 were predicted using three differentiated parameter settings (baseline scenario, low-carbon scenario, and enhanced low-carbon scenario). The result is shown in Figure 3.

In the baseline scenario, the CO₂ emissions from energy consumption in Jiangsu Province in 2025 will be 791.709 million tons, and in 2030, the CO₂ emissions will be 796.828 million tons. Under the low-carbon scenario, the CO₂ emissions from energy consumption in Jiangsu Province in 2025 and 2030 will be 786.876 million tons and 786.355 million tons, respectively. Compared with the baseline scenario, the CO₂ emissions from energy consumption in 2030 will decrease by 10.473 million tons, with a reduction rate of 1.314%. Under the strengthened low-carbon scenario, the CO₂ emissions from energy consumption in Jiangsu Province will be 783.092 million tons in 2025 and 772.293 million tons in 2030. Compared to the baseline scenario, CO₂ emissions will decrease by 24.535 million tons in 2030, with a reduction rate of 3.120%.

3.6. Suggestions and Limitations for Carbon Reduction

(1)

Strengthen the guidance of policies and institutional systems. In the short-term phase (2025–2026), the Jiangsu Provincial Government may actively launch special subsidies for the “coal-to-electricity” transition, promote technological transformation in the iron and steel as well as chemical industries, and establish a two-level (provincial-municipal) carbon emission monitoring platform. Policies shall be formulated to reduce the proportion of coal consumption in the province by 1.5% annually, with the proportion dropping to below 55% by 2026. In the medium-term phase (2027–2030), efforts shall be intensified to construct gigawatt-scale offshore wind farms (e.g., in Yancheng and Lianyungang), implement a “carbon labeling” system, and prioritize the procurement of low-carbon products. The share of renewable energy installed capacity shall reach 40%, and the carbon productivity of industrial enterprises above the designated size shall increase by 20%. In the long-term phase (2031-), the Provincial Government and the Department of Ecology and Environment may actively build a “carbon trading + carbon tax” mechanism, develop pilot “zero-carbon parks,” and achieve a stable decline in carbon emissions from energy consumption after reaching the peak, with net carbon emissions decreasing by 10% compared with the peak level.

(2)

Optimize the energy-consumption structure. Under the baseline scenario, fossil fuels are expected to account for 75.630% of Jiangsu’s total energy demand in 2030, whereas the enhanced low-carbon scenario reduces this share to 69.550%. To accelerate this transition, Jiangsu should aggressively expand the deployment of renewables and other clean-energy sources while systematically phasing out coal and other high-carbon fuels. Consequently, Jiangsu must substantially scale up the deployment of renewable and clean-energy sources to diminish its reliance on coal and other polluting fuels, accelerate the optimization and upgrading of its energy-consumption mix, and promptly phase out or replace backward capacities characterized by high energy intensity and high pollution [52,53].

(3)

Scale up technology acquisition and innovation while aligning market and policy mechanisms. First, it is recommended to establish a Special Scientific Research Fund for Carbon Reduction Technologies in Jiangsu Province, with an annual investment of CNY 1.500 billion to support joint research and development efforts between universities and enterprises. The fund shall focus on achieving breakthroughs in key technologies such as “low-carbon metallurgy” and “industrial waste heat recovery”, and mandate that the achievements of the funded projects must be applied in industrialization within 2 years. Second, establish a Carbon Reduction Technology Trading Platform, which connects with the Jiangsu National Independent Innovation Demonstration Zone. Implement a “value-added tax (VAT) immediate refund upon collection” policy for carbon reduction technologies purchased by enterprises, with an annual maximum refund limit of CNY 50 million per enterprise. Meanwhile, propose the Special Program for Carbon Reduction Technology Talents: provide a CNY 5 million settlement subsidy for introduced high-level overseas talents (e.g., academician teams in the new energy field). Furthermore, the province must deepen cooperative exchanges with other regions and organisations on CO₂ mitigation technologies to secure robust technical support for its emission-reduction goals [54,55].

(4)

Accelerate the diffusion of energy-saving and carbon-mitigation technologies. The provincial government should publish an annually updated catalogue of proven, scalable low-carbon technologies and mandate that large enterprises, especially those with high-energy and high-emission properties, act as first movers in its promotion and deployment [7,56,57]. Fiscal incentives and concessional finance should be directed toward research and development of next-generation low-carbon products and processes, while competitive grants establish demonstration plants and carbon-saving pioneer posts to showcase best practices. To close current market gaps, Jiangsu must also pioneer domestic carbon-sink trading and develop carbon-futures and other derivative instruments, thereby deepening the province’s carbon-finance ecosystem.

Moreover, the limitations of this study are as follows: (i) The lack of statistical data on some subdivided energy varieties (e.g., biomass energy) leads to incomplete coverage of all energy consumption within the accounting scope. (ii) The availability of carbon emission data for cities in Jiangsu Province is insufficient, making it impossible to conduct city-level scale analysis. (iii) The extended STIRPAT model does not incorporate spatial spillover effects (e.g., the impact of carbon emissions from neighboring provinces on Jiangsu), which may underestimate the inter-regional correlation effects. (iv) The scenario analysis fails to consider the impact of “extreme events” (e.g., energy price fluctuations, epidemics).

4. Conclusions

(1)

Jiangsu’s energy-related CO₂ emissions surged from 215.224 million tons in 2000 to 783.943 million tons in 2023, averaging 549.963 million tons. Since the 13th Five-Year Plan period, the effects of emission reduction policies have initially emerged. Coal consumption remains the main source of carbon emissions (accounting for over 60%), while the proportion of natural gas and renewable energy has increased slowly. Therefore, the transformation of energy structures is still the key measure for emission reduction.

(2)

Over the same period, the decoupling relationship between energy-related CO₂ emissions and economic growth was predominantly weak decoupling (91.304%), with only a minority of years exhibiting expansive coupling (8.696%). The weak decoupling state indicates that “economic growth still relies on the drive of energy consumption”, and it is necessary to achieve “strong decoupling” through technological innovation and structural optimization.

(3)

Under the baseline scenario, Jiangsu’s energy-related CO₂ emissions are projected to reach 796.828 million tons in 2030. Under the low-carbon scenario, CO₂ emissions decline to 786.355 million tons, and under the enhanced low-carbon scenario, CO₂ emissions fall further to 772.293 million tons.

(4)

Jiangsu Province needs to promote carbon peaking with “energy structure transformation as the core, technological innovation as the support, and policy coordination as the guarantee”. The research conclusions can provide references for economically developed provinces in eastern China, such as Zhejiang and Guangdong. Meanwhile, in the future, “satellite remote sensing data” can be combined to supplement municipal-scale data and improve spatial accuracy; the “spatial STIRPAT model” can be introduced to analyze regional spillover effects; or the “system dynamics model” can be adopted to integrate the impact of extreme events, so as to optimize scenario prediction.