Introduction

Global warming has emerged as an existential threat to humanity, with its accelerating impacts undermining the foundation of sustainable development. Alarmingly, 2024 was documented as the warmest year in observational history, reaching ~1.55 °C above pre-industrial level (World Meteorological Organization, 2025) and surpassing the Paris Agreement’s 1.5 °C long-term temperature goal—a threshold breach which IPCC AR6 models had originally projected would not occur before the 2030s. Carbon emissions are the primary cause of global warming; therefore, most countries have set targets to achieve carbon neutrality by the mid-21st century. However, resources for reducing carbon emissions are still very limited, with G20 countries only allocating 6% of their stimulus spending directed toward this area (Nahm et al., 2022). Therefore, to achieve carbon neutrality goals, it is urgent to establish an evaluation framework for carbon reduction efforts, which directly affects the sustainable development of humanity in the future.

Compared to the carbon reduction trends already observed in advanced economies, China’s carbon emissions increased by 0.565 Gt, reaching a total of 12.6 Gt in 2023, accounting for 33.7% of the global total (International Energy Agency, 2024). Clearly, China’s role in emission reduction is pivotal to global climate change mitigation, as evidenced by its 2020 commitment to achieve carbon neutrality by 2060. To achieve carbon neutrality, China must accelerate carbon intensity reductions. Although its carbon intensity had decreased by 48.4% from 2005 to 2020 (Liu et al., 2022), current projections indicate that the carbon intensity of end-use energy must decline by at least 60% by 2050 (Duan et al., 2021). Additionally, annual negative carbon emissions of 0.01–2.91 Gt will be required between 2050 and 2060 (Deng et al., 2024). Therefore, evaluating China’s transition to carbon neutrality is crucial for advancing global climate governance.

Accounting for 27.7% of global fossil fuel consumption (IEA, 2024), China is a major contributor to worldwide carbon emissions, given that fossil fuel combustion is the primary source of CO2. To reduce dependence on fossil fuels, the country considered developing renewable energy as a key measure to address climate change and achieve carbon reduction (Qin et al., 2024). A growing body of empirical evidence indicates that higher renewable energy consumption significantly reduces China’s carbon emission intensity (Yao et al., 2019; Yu et al., 2020; Zheng et al., 2021). The Chinese government has actively promoted renewable energy development through strategic policy frameworks and targeted investments, driving a remarkable increase in the share of wind and solar power in total installed capacity from 7.4% in 2013 to 29.5% in 2022 (National Bureau of Statistics of China, 2023). China’s installed capacity of renewable energy increased by 117 gigawatts (GW) in 2022 compared with that in 2021, accounting for 39.7% of the global increase of 295 GW (International Renewable Energy Agency, 2023). However, scholars have also examined challenges in China’s renewable energy development. For instance, wind and solar power generation efficiency is constrained by the significant seasonality, volatility, and instability of local climate conditions, resulting in relatively low output levels (Matamala and Feijoo, 2021). Due to differences in resource endowments, a geographical mismatch exists between electricity demand centers and renewable energy generation centers (Jia et al., 2022). Therefore, one objective of this study is to evaluate regional disparities in renewable energy development across China, with the aim of optimizing the pathway toward carbon neutrality.

In addition to developing renewable energy, ecosystem carbon sinks, especially forest carbon sink, are considered another important way to achieve carbon neutrality (Mo et al., 2023). The potential forest land in the world will reach 1.7–1.8 billion hectares, and there is room for 0.9 billion hectares of canopy cover (Bastin et al., 2019). The global potential for climate change mitigation through afforestation is 1.5 billion tons CO2 / year (Doelman et al., 2020). China is the fastest-growing country in afforestation worldwide. The annual carbon sink potential of afforestation in China is projected to reach 104 million metric tons of CO2 during the period from 2015 to 2060 (Cai et al., 2022). In addition, scholars have noted that increasing carbon sinks through afforestation is more cost-effective than other approaches to achieving carbon neutrality (Xu et al., 2022; Cai et al., 2023). However, there are regional disparities in China’s afforestation (Cao et al., 2023), and the impact of regional heterogeneity must be considered when evaluating afforestation effectiveness (Ge et al., 2023). Thus, this study integrates regional afforestation disparities to quantify the efficiency of China’s carbon neutrality transition, with a focus on forest carbon sink contributions.

The Chinese government’s long-term financial commitment to energy conservation and environmental protection, including policy support and funding, plays a crucial role in supporting the rapid development of renewable energy and afforestation. Figure 1 compares China’s finance expenditure on energy conservation and environmental protection (EECEP), wind and solar power generation, and afforestation area from 2013 to 2022, with year-on-year change data for each category. In 2020, expenditure declines caused by COVID-19 reduced afforestation and renewable energy utilization, aligning with the trend observed by Matthews and Wynes (2022) in G20 countries.

Fig. 1
Fig. 1
Full size image

Comparison of China’s expenditure on energy conservation and environmental protection, wind and solar power generation, and afforestation with Year-on-Year Changes.

Therefore, this study initially assumes that achieving carbon neutrality requires governments to allocate resources commensurate with their emission levels, thereby meeting specific reduction targets. Based on this initial assumption, this study develops a carbon neutrality transition evaluation framework to evaluate the Chinese government efforts toward carbon neutrality goals. First, using CO2 emissions as the linkage, this study establishes an interaction mechanism between emission sources and carbon reduction subsystems, then applies a two-stage dynamic undesirable non-radial directional distance function (NDDF) to evaluate provincial carbon neutrality transition efficiency (CNTE) in China. Second, by accounting for regional disparities in resource endowments using land availability and expenditure as inputs, this study quantifies the potential for wind/solar power generation and afforested carbon sinks, thereby formulating policy recommendations for optimizing China’s carbon neutrality transition pathways.

The rest of this study runs as follows. Section “Literature review” provides a literature review. Section “Research and data methodology” shows the research and data methodology. Section “Empirical analysis” presents the empirical study. Section “Discussions and policy recommendations” offers discussions and policy recommendations. Section “Conclusion and implications” summarizes conclusions and implications.

Literature review

Progress in carbon neutrality transition evaluation

Accurate carbon neutrality evaluation is a prerequisite for meeting carbon neutrality targets (Chen et al., 2024). Current studies show that carbon neutrality evaluation focuses on regional emission accounting and efficiency analysis. As a key indicator for evaluating environmental quality and carbon neutrality, carbon emission efficiency (CEE) lays an important fundamental for evaluation frameworks (Dong et al., 2022). Given that the precision of carbon emission data directly determines the effectiveness of global climate governance, it constitutes a core basis for formulating emission reduction policies and setting quantifiable targets (Li et al., 2024). However, due to variations in data availability, measurement tools, and methodological approaches, the scope of emission accounting exhibits significant heterogeneity across different studies (Huovila et al., 2022). Additionally, regional carbon neutrality levels are typically characterized by the ratio or difference between carbon emissions and carbon sinks (Cai et al., 2023; Zhan et al., 2024; Li et al., 2025). This metric effectively reflects the potential of regional carbon sequestration capacity to offset anthropogenic emissions (Li et al., 2024).

In recent years, scholars have also developed multidimensional frameworks to more systematically advance the assessment of carbon neutrality progress (Niu et al., 2021; Yang et al., 2022; Chun et al., 2023). For instance, Xi et al. (2024) highlighted the importance of embodied carbon flows and transregional carbon sink services and consequently proposed an integrated assessment framework that incorporates both. Furthermore, as the role of non-fossil alternatives in the energy supply side becomes increasingly prominent, Chen et al. (2025) advocated for using clean energy outputs, such as hydropower, wind power, nuclear power, and solar energy as core indicators, thereby extending carbon neutrality progress assessment from emission control to energy system transformation.

As a natural solution, afforestation is considered a cost-effective and readily available option for mitigating climate change (Alkama et al., 2016; Rogelj et al., 2018; Duffy et al., 2020; Doelman et al., 2020). Afforestation and reforestation contribute to climate change mitigation via enhanced carbon storage in biomass and soils (Wang et al., 2025). As the duration of afforestation and reforestation increases, the net ecosystem exchange of forests shows a declining trend, indicating that forests restored through ecological recovery can rapidly become carbon sinks (He et al., 2024). Simultaneously, coastal wetlands and marine carbon sinks show growing potential for absorbing CO2 (Howard et al., 2017; Iversen et al., 2023). Nevertheless, the annual mean ocean carbon sink varies primarily caused by changes in the sources and sinks of natural CO2, with a lesser role for variations in atmospheric CO2 growth rates impacting the uptake of anthropogenic CO2 (Gruber et al., 2023). Compared to enhancing ocean carbon sinks, afforestation offers greater implementation feasibility. For instance, Xu et al. (2023) formulated an afforestation roadmap aligned with China’s 2060 carbon neutrality goal. Zhai et al. (2023) emphasized that China’s Three-North Shelterbelt Program, the world’s largest afforestation initiative, primarily combats desertification while enhancing forest carbon sequestration capacity.

In the field of energy transition, the shift to renewable energy has become a critical strategy in response to the depletion of fossil fuels and growing global energy demand (Gunathilake et al., 2024). Renewable power generation can significantly reduce carbon emissions and improve environmental quality (Bilgili et al., 2025). The power sector is the primary source of CO2 emissions in China and represents a pivotal domain for achieving carbon emission reduction targets (Sun et al., 2022; Wei et al., 2024). Increasing the share of renewable energy in the power generation mix can effectively reduce dependence on fossil fuels, thereby significantly decreasing greenhouse gas emissions (Paraschiv et al., 2023). Li et al.(2022) reported that China’s annual wind power generation potential ranges between 12,900 and 15,000 TWh, while its solar energy potential ranges from 3100 to 5200 TWh. The theoretical maximum capacity could triple the total national electricity consumption recorded in 2019, and this capacity is sufficient to support the goal of carbon neutrality by 2060. Although both wind and solar power have continuously increased their installed capacity and proportion of electricity generation in the power system (Liu et al. 2022, 2024), their actual contribution to emission reduction still lacks systematic quantitative assessment. To address this gap, Teng et al. (2022) introduced the Carbon Emission Reduction Capacity (CERC) indicator using the IPCC 2006 carbon emission factors. However, this indicator assumes lossless year-round operation of wind and solar power generators year-round, overlooking the impact of extreme weather on power generation facilities—a simplification that requires optimization using real-world power generation data.

In addition, nuclear energy continues to play an important role in the global energy transition due to its near-zero emissions, mature technology, and reliable power supply (Carrara, 2020). However, due to national security concerns, it is difficult to obtain consistent panel data on nuclear power generation. Furthermore, carbon capture, utilization, and storage (CCUS) technology is regarded as a key support for achieving carbon peak and carbon neutrality goals (Davoodi et al., 2024). Direct Air Capture (DAC) technology directly separates carbon dioxide from the atmosphere through chemical or physical methods, providing an innovative solution for deep decarbonization (Jiang et al., 2023). However, technologies such as CCUS and DAC remain constrained by high costs and storage safety concerns (Kar et al., 2025). Moreover, Chu et al. (2024) conducted a comprehensive assessment of three mitigation measure categories—energy system upgrades, biotechnology applications, and CCUS—within national policy frameworks, highlighting that their selection requires balancing economic conditions and natural resource endowments across different countries.

Building on existing literature, this study integrates carbon reduction systems into carbon emission evaluation frameworks to develop a comprehensive evaluation model for the transition to carbon neutrality, thus enabling the quantification of government-led decarbonization efforts. Given the disparities in factor endowments across Chinese regions, the afforestation area and the modified CERC will be proposed as the key indicators for tracking China’s carbon neutrality transition.

Evolution of DEA method

Data envelopment analysis (DEA) is an efficiency evaluation method traceable to Farrell (1957) and applicable to systems with multiple inputs and outputs. One early application of the Charnes-Cooper-Rhodes (CCR) (Charnes et al., 1978) and Banker-Charnes-Cooper (BCC) (Banker et al., 1984) models in evaluating China’s energy performance can be traced back to Hu and Wang (2006), which took labor, capital, energy consumption and total sown area as multiple input variables, GDP as a single output variable, and China’s provinces as decision-making units (DMUs). Subsequently, Wei et al. (2009) and Chang and Hu (2010) established a basic framework for evaluating China’s energy efficiency, which takes labor, capital, and energy consumption as input variables and GDP as the output variable. With increasing academic attention to CO2 emissions, Yeh et al. (2010) employed the CCR and BCC models to assess regional energy efficiency in China, explicitly treating CO2 as an undesirable output.

However, both the CCR and BCC models are linear programming approaches grounded in the underlying assumption of proportional relationships between inputs and outputs. This inherent limitation may lead to measurement errors when analyzing non-proportional scale changes or systems generating undesirable outputs (e.g., CO2 emissions). Tone (2001) proposed the Slacks-Based Measure (SBM) model, which solves the problem when an input or output is not being adjusted proportionally to achieve optimal efficiency. Subsequently, Färe and Grosskopf (2010) proposed a NDDF by integrating the SBM model with the directional distance function (Chung et al., 1997), providing more reasonable and accurate results for evaluating undesirable outputs.

NDDF has been widely applied in evaluating CO2 emission performance of energy systems. For instance, Wang et al. (2013) used NDDF to analyze improvements in energy efficiency and productivity, finding that the main driver of energy productivity growth in China was technological change rather than efficiency enhancement. Li and Lin (2015) employed a metafrontier NDDF framework to evaluate China’s regional energy performance, finding that eastern China exhibited the highest efficiency, followed by western and central China.

The aforementioned CCR, BCC, SBM, and NDDF models are static DEA frameworks, which limits their applicability for cross-period efficiency comparisons and trend change analysis. As emphasized by Kao (2013), dynamic efficiency analysis is essential when data allow, since system efficiency is defined as a linear combination of period efficiencies in dynamic systems. Thus, ignoring this dynamic nature leads to efficiency overestimation. Early dynamic DEA models applied to energy efficiency evaluation primarily include the DEA window analysis model (Charnes and Cooper, 1984) and the Malmquist index (Caves et al., 1982). However, since both focus on individual time periods and neglect carry-over activities between consecutive time periods, they can only achieve local optimization for each time period independently (Tone and Tsutsui, 2010). To address this limitation, Tone and Tsutsui (2010) proposed a dynamic SBM model that incorporates carry-over activities and enables the measurement of period-specific efficiencies via long-term optimization across the entire timeframe. The dynamic SBM model has been widely employed to evaluate regional energy efficiency and carbon emissions performance, as evidenced by studies of OECD countries (Guo et al., 2017), APEC economies (Ke, 2017), and Belt and Road economies (Mills et al., 2021). Teng et al. (2021) developed a dynamic SBM framework that incorporating afforestation as an exogenous variable to evaluate China’s energy efficiency performance, thereby emphasizing its role as a critical climate change mitigation strategy.

By allowing flexible quantification of slack variables for multiple input-output dimensions, the NDDF model improves evaluation accuracy. This capability has driven recent integrations of the DDF framework with dynamic analysis, which assess multi-variable impacts on energy efficiency. For instance, Teng et al. (2022) examined the role of renewable energy deployment in enhancing energy efficiency by constructing a dynamic NDDF framework. This framework incorporated the CERC of wind and solar power as an output variable to evaluate China’s energy efficiency performance. Teng et al. (2023) incorporated the annual increment of live wood stock as a desirable output and proposed that regional disparities in forest carbon sequestration across China should be accounted for in the evaluation of energy efficiency. Xie et al. (2023) incorporated financial expenditure as an input variable to evaluate China’s energy efficiency, emphasizing that increased investment in energy conservation and environmental protection could improve energy efficiency. The limitation lies in the fact that the above study only quantified the improvement potential of individual variables for carbon reduction, failing to reflect the overall performance of these measures.

This limitation, stemming from sub-technologies regarded as a “black box”, can be evaluated using a network DEA framework to assess their efficiency at different stages. Färe et al. (2007) proposed network DEA, which applied sub-technologies to examine the impact of intermediate products on the production process, avoiding the treatment of such products as “black box” excluded from evaluation. Following Färe et al. (2007), Tone and Tsutsui (2009) set up a weighted SBM-DEA model, which considered intermediate products between departments as linkages and used it to find optimal solutions. Subsequently, Tone and Tsutsui (2014) proposed the weighted SBM dynamic network DEA model, which introduced carry-over variables to link time periods and enable dynamic analysis of network structures over time.

Many scholars have applied the two-stage network DEA method to inter-sectoral efficiency evaluation. For instance, Wang et al. (2022) constructed a two-stage SBM DEA model and analyzed the linkage between ports and regional economic systems, demonstrating the contribution of ports on the regional economy. For applications of the SBM dynamic network DEA model, Lin et al. (2021) employed the undesirable outputs of first-stage energy consumption as inputs for second-stage healthcare efficiency evaluation, concluding that China’s energy consumption efficiency is superior to its healthcare efficiency. Wang et al. (2021) used pollutants to link the evaluation of production and treatment performance of China’s industrial sector, finding that production efficiency outperformed treatment efficiency. The two-stage dynamic NDDF has also been applied in recent research. Teng et al. (2024) evaluated China’s energy performance in the first stage and established CO2 emissions as a linking variable to evaluate performance in coping with extreme weather in the second stage, finding that the efficiency in coping with extreme weather was far lower than energy efficiency. Jiang et al. (2025) used energy as a link to evaluate energy extraction efficiency in the first stage and energy utilization efficiency in the second stage, and obtained results for 35 key fossil energy-producing Belt and Road Initiative (BRI) countries.

The key difference lies in this study’s application of a two-stage dynamic NDDF to evaluate the efficiency of China’s carbon neutrality transition. Following traditional energy efficiency evaluation frameworks, CO2 emissions are designated as an undesirable output in the first stage of the carbon emission subsystem and are linked to the second stage. Based on the review in Section “Progress in carbon neutrality transition evaluation”, the second stage is defined as the carbon reduction subsystem, with afforestation area and CERC as its output variables. Input variables encompass land use (Teng et al., 2023) and financial expenditure (Xie et al., 2023), in addition to first-stage CO2 emissions, to reflect governmental resource allocation and carbon reduction initiatives. Based on the efficiency values and the improvement potential of variables across the two stages, this study proposes policy recommendations to optimize system transition pathways for carbon neutrality in China.

Research and data methodology

In the first stage, CEE is evaluated as a key metric for the carbon emission system’s performance. According to the China Statistical Yearbook, year-end population (i.e., Population) and total regional energy consumption (i.e., Energy) are set as input variables, while gross regional product (i.e., GDP) is designated as the desirable output variable. Furthermore, according to the China Energy Statistical Yearbook, CO2 emissions in the region are set as the undesirable output variable, calculated using the IPCC 2006 emission factors based on fossil fuel consumption (including coal, coke, crude oil, gasoline, kerosene, diesel oil, and natural gas).

In the second stage, Carbon Reduction Efficiency (CRE) is evaluated as a key metric for the carbon reduction system’s performance. The input variables comprise CO2 emissions (a linkage variable transferred from the first stage), along with EECEP and land use (i.e., Land). The afforestation area is defined as a desirable output variable based on EECEP and land input. This metric can be directly obtained from the China Statistical Yearbook, which also provides input data for both EECEP and land-related variables. According to Teng et al. (2022), CERC is defined as an additional desirable-output variable in the second stage. Following this definition, this study calculates CERC based on actual wind and solar power generation using the following formula:

$${CERC}=\frac{\left({E}_{{wind}}+{E}_{{solar}}\right)* {{CF}}_{{coal}}}{{{E}F}_{c{oal}}}$$
(1)

Ewind and Esolar represent wind and solar power generation, respectively, with data sourced from the China Electric Power Statistical Yearbook. CFcoal denotes the emission coefficient of coal-fired power generation, which is 1.08 kg CO2/kWh, calculated based on the IPCC 2006 CO₂ emission factors, the assumed 33% power generation efficiency, and the lower heating value of lignite. EFcoal represents the standard coal CO₂ emission factors from IPCC guidelines, which is 2640 kg CO2/tce.

According to the China Statistical Yearbook, the total fixed-asset investment in the whole region (i.e., Capital) serves as a carry-over variable to link period t with period t + 1 across the two stages. Figure 2 presents the framework for evaluating CNTE.

Fig. 2
Fig. 2
Full size image

Framework for evaluating carbon neutrality transition efficiency.

NDDF provides more accurate evaluation results when dealing with multiple types of input and output variables. Thus, we modify the traditional NDDF (Färe and Grosskopf, 2010) to develop an enhanced version, which is then integrated with the dynamic network DEA framework (Tone and Tsutsui, 2014) to construct a two-stage dynamic NDDF for evaluation.

Suppose there are t time periods \(\left(t=1,\ldots ,T\right)\). In each time period, there are two stages, including carbon emission stage and carbon reduction stage. The former has m inputs\(\,{x}_{{ij}}^{t}\left(i=1,\ldots ,m\right)\) to generate D intermediate products \({z}_{{dj}}^{t}\left(d=1,\ldots ,D\right)\), K desirable outputs \({q}_{{kj}}^{t}\left(k=1,\ldots ,K\right)\) and L undesirable outputs \({m}_{{lj}}^{t}\left(l=1,\ldots ,L\right)\). The latter stage produces S desirable outputs \({y}_{{rj}}^{t}\left(r=1,\ldots ,s\right)\) and O undesirable outputs \({p}_{{oj}}^{t}\left(o=1,\ldots ,O\right)\) from D intermediate products \({z}_{{dj}}^{t}\left(d=1,\ldots ,D\right)\) and G inputs \({w}_{{gj}}^{t}\left(g=1,\ldots ,G\right)\), \({{\rm{where}}\;{c}}_{{hj}}^{{d}^{t-1}}\left(h=1,\ldots ,H\right)\) is a carry-over variable.

Equations (1)–(4) are used to calculate the overall efficiency of CNTE and the efficiency of each stage, as shown below. The objective function for the overall efficiency of CNTE is defined as follows:

$${\rm{Max}}\mathop{\sum }\limits_{t=1}^{T}{\gamma }_{t}({w}_{1}^{t}{\theta }_{1}^{t}+{w}_{2}^{t}{\theta }_{2}^{t})$$
(2)

Subject to:

$$\begin{array}{ll}{\rm{Carbon}}\;{\rm{emission}}\;{\rm{stage}}\qquad{\rm{Carbon}}\;{\rm{reduction}}\;{\rm{stage}}\\\mathop{\sum }\limits_{j}^{n}{\lambda }_{j}^{t}{x}_{{ij}}^{t}\le {\theta }_{1}^{t}{x}_{{ip}}^{t}\forall i,\forall t\qquad\;\;\mathop{\sum }\limits_{j}^{n}{\mu }_{j}^{t}{z}_{{dj}}^{t}\le {\theta }_{2}^{t}{z}_{{dp}}^{t}\forall d,\forall t\\\mathop{\sum }\limits_{j}^{n}{\lambda }_{j}^{t}{z}_{{dj}}^{t}\le {\theta }_{1}^{t}{z}_{{dp}}^{t}\forall d,\forall t\qquad\;\;\mathop{\sum }\limits_{j}^{n}{\mu }_{j}^{t}{p}_{{oj}}^{t}\ge {\theta }_{2}^{t}{p}_{{op}}^{t}\forall s,\forall t\\\,\mathop{\sum }\limits_{j}^{n}{\lambda }_{j}^{t}{q}_{{kj}}^{t}\ge {\theta }_{1}^{t}{q}_{{kp}}^{t}\forall k,\forall t\qquad\;\;\mathop{\sum }\limits_{j}^{n}{\mu }_{j}^{t}{y}_{{sj}}^{t}\ge {\theta }_{2}^{t}{y}_{{sp}}^{t}\forall s,\forall t\\\mathop{\sum }\limits_{j}^{n}{\lambda }_{j}^{t}{m}_{{lj}}^{t}\le {\theta }_{1}^{t}{m}_{{lp}}^{t}\forall k,\forall t\qquad\;\,\mathop{\sum }\limits_{j}^{n}{\mu }_{j}^{t}{w}_{{gj}}^{t}\le {\theta }_{2}^{t}{w}_{{gp}}^{t}\forall g,\forall t\\\mathop{\sum }\limits_{j}^{n}{\lambda }_{j}^{t}\le 1\qquad\qquad\qquad\qquad\;\mathop{\sum }\limits_{j}^{n}{\mu }_{j}^{t}=1\\{\lambda }_{j}^{t}\ge 0\,\forall j,\forall t\qquad\qquad\qquad\quad\;{\mu }_{j}^{t}\ge 0\,\forall j,\forall t\end{array}$$
(3)

Link of the two stages

$$\mathop{\sum }\limits_{j=1}^{n}{\lambda }_{j}^{t}{Z}_{{dj}}^{t}=\mathop{\sum }\limits_{j=1}^{n}{\mu }_{j}^{t}{Z}_{{dj}}^{t}\;\forall d,\forall t$$
(4)

Link of the two periods

$$\mathop{\sum }\limits_{j=1}^{n}{\lambda }_{j}^{t-1}{c}_{{hj}}^{t}=\mathop{\sum }\limits_{j=1}^{n}{\lambda }_{j}^{t}{c}_{{hj}}^{t}\;\forall h,\forall t$$
(5)

Among them \({\gamma }_{t}\) is the weight assigned to period t, and \({w}_{1}^{t}\) and \({w}_{2}^{t}\) are the weights assigned to the carbon emission stage and the carbon reduction stage in time period t, respectively. Therefore, for each time period t \({w}_{1}^{t},{w}_{2}^{t},{\gamma }_{t}\ge 1\) and \(\mathop{\sum }\nolimits_{t=1}^{T}{\gamma }_{t}=1.\) For the sake of analytical clarity and grounded in the research motivation and theoretical justification, both the period weights and stage weights are assumed to be equal to 1.

Based on the above methods, this study evaluates the overall efficiency defined as CNTE and defines the division efficiencies as CEE and CRE of 30 provinces in China from 2013 to 2022. Table 1 reports the statistical description of variables.

Table 1 Descriptive statistics of variables.
Full size table

Empirical analysis

Efficiency results

This study analyzes 30 provinces in China as DMUs to evaluate their CNTE during 2013–2022. The overall efficiency, calculated using a two-stage dynamic NDDF, is defined as CNTE. It can be decomposed into two component efficiencies: CEE in the first stage and CRE in the second stage. Using CO2 emissions as the linkage, this approach establishes a causal relationship between emission systems and reduction systems, and calculates CNTE as the product of CEE and CRE. Table 2 reports the average CNTE and the changes in CEE and CRE during the evaluation period. All efficiency values range from 0 to 1, and DMUs achieving an efficiency of 1 constitute the benchmarks for calculating the efficiency of other DMUs. Figures 3 and 4 present regional comparisons of the 10-year averages for CEE and CRE, respectively. The main findings are summarized as follows.

  1. i.

    The CNTE shows an initial increase from 2013 to 2017, followed by a decrease from 2018 to 2022. This trend is basically consistent with the finance expenditure increase rate trend reported in Fig. 1. After decomposing the CNTE, the CEE averaged ~0.7 from 2013 to 2019. However, its average for 2020–2022 dropped to 0.68, which is lower than the 2013–2019 average level. The CRE rose from 0.73 in 2013 to 0.77 in 2019, before declining to 0.73 in 2022, thereby returning to its 2013 level. In contrast, the impact of CRE on CNTE is greater than that of CEE.

  2. ii.

    The CEE of DMUs in eastern region (e.g., Beijing, Fujian, Guangdong, Hainan, Jiangsu, Shanghai, and Zhejiang) is generally better than those in CRE. These DMUs share a common characteristic: lower energy and CO2 emission intensities per unit of GDP, yet their land resource endowments are less favorable than those of central and western DMUs. In contrast, DMUs in the central and western regions (e.g., Chongqing, Gansu, Guangxi, Guizhou, Hunan, Jiangxi, Qinghai, Shaanxi, Xinjiang and Yunnan) demonstrate significant advantages in CRE.

  3. iii.

    Some provinces exhibit marked contrasts between their CEE and CRE values. For certain DMUs (e.g., Chongqing, Gansu, Guangxi, Hebei, Hunan, and Qinghai), the low CNTE is primarily due to a decline in CEE. In contrast, for others (e.g., Guangdong, Jiangsu, and Zhejiang), low CNTE results from a deterioration in CRE. Meanwhile, some provinces (e.g., Hainan, Heilongjiang, Henan, Shandong and Sichuan) show low performance in both CEE and CRE, and thus have low CNTE. Resource endowment and government input are the key factors contributing to these regional differences, which will be further analyzed in Section “Discussions and policy recommendations” of this study.

Table 2 Regional CNTE and change of CEE & CER from 2013 to 2022.
Full size table
Fig. 3
Fig. 3
Full size image

Regional comparison of carbon emission efficiency.

Fig. 4
Fig. 4
Full size image

Regional comparison of carbon reduction efficiency.

Robustness analysis

This subsection discusses the robustness analysis by further examining the consistency of efficiency results across different models through correlation tests.

First, based on the input and output framework presented in Fig. 2, the robustness of the DDF and SBM models is analyzed. Next, building on the economic production theory which emphasizes labor as a fundamental and essential input in production, this study conducts a robustness analysis by replacing the original population input with labor input.

H0: The Carbon Neutrality Transition Efficiency (CNTE), Carbon Emission Efficiency (CEE), and Carbon Reduction Efficiency (CRE) derived from the DDF model are positively correlated.

H1: The CNTE, CEE, and CRE derived from the SBM model are positively correlated.

H2: The CNTE, CEE, and CRE before and after substituting the input variable are positively correlated.

Under the DDF model framework, Table 3 presents the correlation test results of CNTE, CEE, and CRE from 2013 to 2022. Overall, the correlation coefficients among CNTE, CEE, and CRE all exceed 0.5, and the p-values are less than 0.05, indicating a high degree of correlation. Therefore, the null hypothesis (H0) is accepted.

Table 3 Correlation tests of CNTE, CEE, and CRE from 2013 to 2022 under the DDF model framework.
Full size table

Under the SBM model framework, Table 4 shows the correlation test results of CNTE, CEE, and CRE from 2013 to 2022. Overall, the correlation coefficients among CNTE, CEE, and CRE all exceed 0.5, and the p-values are less than 0.05, indicating a strong correlation. Therefore, the alternative hypothesis (H1) is accepted.

Table 4 Correlation Tests of CNTE, CEE, and CRE from 2013 to 2022 under the SBM model framework.
Full size table

A comprehensive review of Tables 3 and 4 clearly indicates that, regardless of whether the DDF model or the SBM model is used, there exists a strong correlation among CNTE, CEE, and CRE across both efficiency evaluation frameworks.

As shown in Table 5, there is no significant difference between the results before and after the substitution of the input variable. Therefore, the hypothesis H2 is accepted.

Table 5 Correlation Tests of CNTE, CEE, and CRE from 2013 to 2022 before and after input variable substitution.
Full size table

Discussions and policy recommendations

The path to improving carbon neutrality transition efficiency

Based on empirical findings, CNTE is synergistically driven by CEE improvement in Stage 1 and CRE improvement in Stage 2. To identify strategic optimization pathways for CNTE, the following analysis focuses on core metrics of these dual drivers.

As shown in the results of Wang et al. (2013) and Li and Lin (2015), DMUs in the central and western regions exhibit higher energy intensity (i.e., energy consumption per unit of GDP), which generates more CO2 emissions and indicates lower energy efficiency. However, existing studies fail to quantify the government’s emission reduction efforts and their actual carbon mitigation effects. To address this gap, this study introduces EECEP as a verifiable indicator for evaluating reduction efforts in the analysis.

Figure 5 shows that DMUs with lower energy intensity do not necessarily have better CNTE, which may be attributed to insufficient government investment in carbon reduction initiatives. Specifically, in the first quadrant —where DMUs exhibit both high energy intensity and a high proportion of EECEP to GDP—most regions have CNTE values exceeding the average by 0.71, such as Ningxia, Inner Mongolia, Gansu, Hebei, and Guizhou. Conversely, in the third quadrant—where DMUs exhibit lower energy intensity and a smaller proportion of EECEP to GDP—most of them perform poorly in terms of CNTE, such as Guangdong, Jiangsu, Zhejiang, Shandong, and Fujian. The above observations indicate that increasing expenditure on carbon reduction is more effective in improving CNTE than reducing energy intensity. Moreover, sustained increases in EECEP represent an effective pathway to enhance carbon reduction efforts and improve CNTE.

Fig. 5
Fig. 5
Full size image

Relationship between energy consumption, EECEP, and CNTE.

China’s primary carbon reduction approaches currently include replacing traditional fossil fuels with renewables like wind and solar power, and expanding ecosystem carbon sinks through afforestation efforts. Figure 6 quantifies the effectiveness of carbon reduction pathways using the ratio of CERC to total CO2 emissions and the proportion of afforested area to total land use, and further correlates these indicators with CNTE. Figure 6 shows that only Ningxia and Hebei perform above average in both afforestation and CERC. This dual strength correlates with their relatively better CNTE performance. Across DMUs in the second and fourth quadrants, the effectiveness of these two carbon reduction pathways varied, and natural resource endowment influenced the pathway selection. For instance, in Xinjiang, a region with extensive desertified land, developing wind and solar power is more effective in reducing carbon emissions than expanding afforestation. However, more DMUs are located in the third quadrant, where the development of the two carbon reduction pathways significantly lags behind, resulting in lower CNTE performance. It is necessary to further explore optimizing carbon emission reduction pathways by analyzing improvement potential in order to provide more targeted policy recommendations.

Fig. 6
Fig. 6
Full size image

Relationship between CERC, afforestation, and CNTE.

Policy recommendations for improving carbon neutrality transition efficiency

As mentioned above, China’s primary strategies for achieving carbon neutrality involve expanding renewable energy (e.g., wind and solar) to replace fossil fuels and enhancing afforestation. In evaluating CNTE, this study derives potential improvement indicators for input-output variables by comparing the performance of actual DMUs against benchmark targets. Tables 6 and 7 report changes in these improvement indicators for CERC and afforestation variables from 2013 to 2022, serving as the basis for the policy recommendations that follow.

Table 6 Evolution of CERC improvement potential from 2013 to 2022 unit:%.
Full size table
Table 7 Evolution of afforestation improvement potential from 2013 to 2022 unit:%.
Full size table

According to Table 6, the improvement potential of CERC shows a declining trend, with its average value dropping from 25.4% in 2013 to 14.8% in 2022. This indicates progress toward balanced renewable energy deployment nationwide. However, there remain significant opportunities for improvement in some DMUs, such as Fujian, Guangdong, Hainan, Heilongjiang; Shandong, Sichuan, and Zhejiang. In contrast, Anhui, Hubei, and Liaoning show a rising trend in their potential for improvement. The aforementioned DMUs are mostly situated in densely populated and industrially concentrated eastern region, where land scarcity has become a critical constraint on developing land-intensive renewable energy. Notably, several DMUs (e.g., Jiangsu and Hebei) with comparable natural resource endowments have achieved significant advancements in renewable energy development, providing valuable operational benchmarks for peer regions.

According to Table 7, whereas the improvement potential in CERC decreases, that of afforestation shows an opposite trend, increasing from 14% in 2013 to 26.7% in 2022. From 2021 to 2022, the growth in improvement potential indicates a significant decline in afforestation’s contribution to carbon neutrality, likely caused by reduced EECEP during the COVID-19 pandemic. In addition, regional disparities in afforestation are widening, particularly in eastern regions such as Jiangsu, Shandong, Shanghai, and Zhejiang, where significant potential for afforestation improvement remains. This reflects how land use restrictions constrain the continuous expansion of afforestation.

Based on the above analysis of potential improvements, we propose the following policy recommendations to enhance the efficiency of China’s transition to carbon neutrality.

  1. i.

    EECEP is the key to supporting carbon emission reduction measures, and sustained growth of EECEP should be ensured. A special fund for carbon neutrality is recommended to be established based on the existing EECEP, with primary objectives including supporting renewable energy generation and enhancing ecosystem carbon sinks.

  2. ii.

    Eastern regions must overcome the structural limitations of their land resource endowments to advance carbon neutrality transition. It is suggested that provinces in the eastern region increase support for Building Integrated Photovoltaics and incorporate afforestation measures into urban development frameworks to promote forest city construction, given their high urbanization rates. Coastal provinces should formulate differentiated support policies and increase subsidies for offshore wind power projects and associated equipment manufacturing—including power distribution, energy storage, and transmission equipment—to mitigate land resource constraints on renewable energy development.

  3. iii.

    Western and central regions should leverage resource endowments to systematically advance carbon neutrality transition and improve decarbonization outcomes. While leveraging land resource advantages to scale up renewable energy and promote afforestation, the central and western regions should focus on converting carbon reduction potential into economic benefits. It is recommended that the central and western regions develop industrial policies aligned with their renewable energy capacity, accelerate workforce training, and facilitate the orderly transfer of energy-intensive industries from the eastern region. This approach will establish a sustainable development path in which economic growth provides long-term financial support for the transition to carbon neutrality.

Conclusion and implications

Main conclusions and contributions

Global warming is accelerating beyond projected rates and poses a critical threat to global sustainability efforts. However, since 2020, the COVID-19 pandemic has compelled countries worldwide to scale back resources allocated to combating global warming. Evaluating governmental carbon reduction initiatives and optimizing pathways toward carbon neutrality represent a critical research problem demanding urgent scholarly attention. While current research predominantly focuses on measuring CEE, systematic methods to evaluate carbon reduction potential remain underdeveloped. To fill the gap in this field, this study attempts to establish a link between carbon emission sources and carbon reduction subsystems and to apply a two-stage dynamic NDDF for the preliminary construction of a carbon neutrality transition evaluation framework.

This study empirically demonstrates that regional disparities in China’s CNTE have exhibited a widening trend since 2020. Government EECEP is a key driver of the aforementioned changes. Regional disparities are further manifested through the eastern region’s efficiency advantage in carbon emission subsystem, whereas central and western regions excel in carbon reduction subsystem. This demonstrates that China’s current wind and solar based renewable energy development and its afforestation path toward carbon neutrality are both heavily influenced by land resource endowment. Based on the above findings, the following are identified as core policy recommendations: (1) establishing a carbon neutrality special fund to ensure financial investment; (2) addressing resource endowment constraints in the eastern region; and (3) transforming western carbon reduction advantages into economic development advantages.

Research limitations and prospects

This study develops an evaluation framework using Chinese provinces as DMUs to assess the efficiency of carbon neutrality transition at the provincial level. Government-level design of input-output indicators lacks explanatory power regarding the technological and industrial pathways necessary to achieve carbon neutrality goals. Future research should refine evaluation frameworks to quantify CNTE across the following dimensions.

  1. i.

    Focus on key technological pathways toward carbon neutrality, using R&D investment and green patents as input indicators to evaluate the effectiveness of CCUS, DAC, BECCS, and hydrogen energy technologies in reducing carbon emissions.

  2. ii.

    Refine the evaluation scope to the city level, focusing on special types of cities (such as resource-based or coastal cities), and evaluate different carbon neutrality paths by accounting for variations in key factors such as ecological carrying capacity and grid connectivity requirements.

  3. iii.

    Given that the industrial sector is the primary source of carbon emissions, its evaluation should incorporate labor, AI technology, and sustainable investment as input indicators to assess low-carbon process reengineering pathways toward carbon neutrality.

  4. iv.

    Future research can employ Tobit regression and the DID method to examine how factors such as Russia-Ukraine war-induced energy price volatility and carbon emissions trading influence CNTE.

However, the optimization of the aforementioned unexplored directions depends on the availability of precise data. Establishing a carbon neutrality evaluation database through national or authoritative research institutions will facilitate future research.