Evaluating China’s New Energy Vehicle Policy Networks: A Social Network Analysis of Policy Coordination and Market Impact

Abstract

Since 2015, China has witnessed a rapid increase in new energy vehicle (NEV) market penetration, achieving global leadership in this sector. This study employs social network analysis (SNA) and Granger causality tests to examine how policy coordination has influenced China’s NEV market development from 2015 to 2023. We evaluated policy coordination using six network metrics: network density, average path length, transitivity, average clustering coefficient, number of components, and size of largest component. Our findings reveal both correlative and causal relationships between policy coordination and market performance. The analysis demonstrated strong positive correlations between network metrics and market performance indicators (ρ = 0.800–0.850, p < 0.01), while Granger causality tests identified significant temporal effects, particularly in the long term (F = 284.051–281,486.748, p < 0.001). Notably, the largest component size shows immediate causal effects on market performance (F = 4.152, p < 0.05). Based on these results, we recommend establishing a multi-level policy coordination system, optimizing the policy network structure with emphasis on core components, implementing dynamic policy adjustment mechanisms considering time-lagged effects, and strengthening collaborative supervision of policy implementation to further advance China’s NEV market development.

1. Introduction

In recent years, global climate change and energy crises have prompted governments worldwide to actively explore sustainable development pathways [1]. In finding and utilizing new energy sources with low-carbon, cost-effective ways have become a key pillar in driving global sustainable economic development [2]. Against this backdrop, new energy vehicles (NEVs), characterized by low emissions, clean energy utilization, and advanced power technologies [3,4], are viewed as a crucial technological approach to address environmental issues and energy security challenges [5].

Since the beginning of the 21st century, the global NEV industry has achieved rapid development. As one of the earliest countries to vigorously promote NEV market development [6], China has achieved remarkable results since designating NEVs as one of seven strategic emerging industries in 2010. According to data from the China Association of Automobile Manufacturers, between 2015 and 2023, China’s monthly average NEV sales increased from 27,400 to 787,400 units, representing a cumulative growth of approximately 27.8 times. Market penetration also rose from 1.34% to 31.45%. Meanwhile, traditional fuel vehicle sales have gradually declined, indicating China’s automotive market is undergoing a structural transition from fuel vehicles to NEVs.

As the world’s largest NEV market, China not only leads in technological research, development, and production scale but has also provided strong support for industry development through a series of policy measures [7]. Since implementing the “Ten Cities, Thousand Vehicles” energy-saving and new energy vehicle program in 2009, the Chinese government has successively introduced numerous policies. These policies span multiple dimensions including financial subsidies, technological innovation incentives, and infrastructure construction [8], reflecting the government’s active intervention in industrial development. However, effective policy implementation depends on coordination among multiple departments, involving numerous institutions such as MIIT, MOF, NDRC, and MOST. With the rapid development of the NEV market, cross-departmental collaboration in policy execution has become increasingly challenging. How this complex policy coordination mechanism affects NEV market development, and what roles and functions various departments play in the policy network, are topics of mutual concern to academics and policymakers.

Despite extensive research on NEV policies, existing literature exhibits three main limitations. First, most studies focus on individual policy effects rather than systematic policy interactions, overlooking the complex interdependencies among different policy instruments and implementing agencies. Second, traditional policy analysis methods often treat policy-making institutions as independent entities, failing to capture the dynamic coordination relationships that characterize China’s NEV policy system. Third, quantitative evaluation of policy coordination’s impact on market performance remains insufficient, particularly regarding the temporal aspects of policy effects.

Social network analysis (SNA) offers unique advantages in addressing these research gaps. First, SNA can effectively visualize and quantify the complex relationships between policy-making institutions, revealing both direct and indirect coordination patterns. Second, network metrics such as density, clustering coefficient, and component size provide standardized measures for evaluating policy coordination efficiency and evolution characteristics. Third, SNA enables systematic examination of the relationship between policy network structure and market performance through correlation and causality analyses. This methodological approach helps understand how policy coordination mechanisms influence market outcomes, providing empirical evidence for optimizing policy frameworks.

Previous applications of SNA in policy research have demonstrated its effectiveness in analyzing policy networks and institutional relationships. However, its application in NEV policy analysis remains limited, particularly regarding the dynamic evolution of policy coordination networks and their market impacts. This study aims to fill this gap by employing SNA to examine China’s NEV policy coordination network, systematically, evaluating its impact on market performance and providing evidence-based recommendations for policy optimization.

2. Literature Review

Governments worldwide are actively developing NEVs against the backdrop of energy environmental impacts and climate change [9]. With the introduction of carbon peak and carbon neutrality goals, NEVs are seen as a crucial lever for achieving a low-carbon economy. To promote NEV adoption, the Chinese government has implemented various policy measures, including the “dual-credit” policy [10], financial subsidies, and infrastructure development planning [11]. These policies have not only increased NEV market penetration but also driven related technological research, development, and application.

China’s NEV policies have multiple dimensions, with core objectives focusing on promoting energy transition, ensuring energy security, and reducing environmental pollution [12]. These objectives are clearly reflected in policy documents such as the “New Energy Vehicle Industry Development Plan (2021–2035)”, demonstrating the national strategic emphasis on the NEV industry. Specifically, policies aim to reduce oil dependency [13] by promoting NEVs, addressing energy security concerns arising from China’s increasing oil import dependence [14,15]. Meanwhile, NEVs significantly reduce carbon dioxide and particulate emissions in the transportation sector [16], providing technical support [17] for mitigating climate change and improving air quality [18]. Additionally, NEV policies are viewed as a key means of promoting green economic transformation [8], supporting related industries to establish an important position in the global NEV market.

NEV policies hold a significant strategic position in China’s energy policy system [7]. They not only form an essential part of energy conservation and emission reduction strategies but also create a synergistic policy system [19] with other energy policies, such as fuel consumption standards, carbon quota management, and new energy credit policies [10]. In the national five-year plans and the medium- to long-term development strategies, NEVs are given priority development status, particularly in the 14th Five-Year Plan, where NEVs are seen as a core lever for promoting green, low-carbon development. Furthermore, the policies highly emphasize infrastructure construction [20], promoting the charging station and hydrogen refueling station deployment to provide convenience for NEV users, while facilitating intelligent and low-carbon development of energy networks. Overall, the strategic importance of NEVs has gained widespread recognition, with their development not only meeting national sustainable development requirements but also providing a “Chinese solution” for global low-carbon economic transformation.

Recent research on NEV adoption and growth has primarily focused on national policies, consumer behavior, and manufacturer strategies [21]. Government policies, particularly financial subsidies, played a crucial role in initial market expansion [22], while comprehensive policy frameworks (including infrastructure construction and R&D support) [23] further drove market growth. At the consumer level, technical performance (such as driving range and charging convenience) [24,25] and operating costs [26] significantly influence purchase intentions, with research focus gradually shifting from acceptance to deeper analysis of usage behavior. For manufacturers, policy support has prompted the optimization of production decisions [27,28] and the balance between profitability and environmental responsibility [29], while subsidy optimization and green production incentive effects have become hot topics.

NEV policy research encompasses multiple dimensions. Research on NEV industry policy tools typically examines supply-side, demand-side, and environmental aspects [30]. Supply-side policy tools mainly focus on enhancing the industry’s technical capabilities and infrastructure construction, including talent cultivation, technical support, capital investment, and infrastructure development [9], aimed at strengthening the overall industrial chain. Demand-side policies promote NEV adoption through market demand stimulation, using tools such as government procurement, market subsidies [31], pilot demonstrations [32], and price guidance. These tools play important roles in reducing market uncertainty. Environmental policies aim to create suitable policy environments for NEV development through regulations, standards, and tax incentives [33], promoting sustainable industry development. Multi-dimensional policy tool coordination is crucial for enhancing policy effectiveness and promoting NEV industry penetration [33].

Despite extensive research on policy tools, studies on coordination relationships between policy-making institutions and their specific impacts on policy effectiveness remain insufficient. Social network analysis (SNA) has recently been introduced to policy research, emerging as a new tool for understanding complex coordination relationships in policy formulation and implementation [34]. SNA can reveal core and peripheral roles of different departments in policy networks while quantifying policy network structural characteristics and evolution patterns through metrics such as density, centrality, and clustering coefficients. SNA offers a powerful lens for understanding the intricate relationships between actors within policy networks, revealing patterns of collaboration, influence, and the diffusion of knowledge and resources [35,36]. Studies leverage various data sources, including policy documents [8,36], patent filings [35], and citation networks [37] to construct and analyze networks. China’s NEV sector is characterized by a complex policy landscape involving a multitude of government agencies. These agencies often engage in collaborative policy-making, as evidenced by co-signer data on policy documents [36]. SNA helps to disentangle these complex relationships, identifying key actors that drive policy formulation and implementation. Researchers have identified several key government agencies that play dominant roles in shaping NEV policy, including the National Development and Reform Commission (NDRC), the Ministry of Science and Technology (MOST), the Ministry of Industry and Information Technology (MIIT), and the Ministry of Finance (MOF) [36]. These agencies leverage their influence to steer policy direction, allocate resources, and foster collaboration within the network.

The evolution of NEV policies in China reflects a strategic approach to addressing industry needs and promoting broader economic goals. Initial policies focused on overcoming technical hurdles, such as limited driving range, while subsequent policies addressed infrastructure challenges, including the availability of charging stations [1,38]. SNA reveals how policy networks adapt to these evolving needs, with new actors emerging and collaborative patterns shifting over time [36]. Financial incentives, including purchase subsidies and tax exemptions, have been crucial in stimulating NEV adoption, leading to significant market expansion [38]. The impressive growth of China’s NEV market is a testament to the effectiveness of its policy coordination and strategic investments, resulting in notable economic benefits. This success can be partly attributed to the government’s robust support for research and development, which has catalyzed technological innovation and fostered a dynamic NEV industry [1,8,35]. The rapid expansion of China’s NEV market has positioned the nation as a global frontrunner in NEV production and sales, which carries significant implications for the country’s industrial competitiveness and energy security. Based on statistical data from the National Bureau of Statistics of China, the production of new energy vehicles (NEVs) in China increased from 8400 units in 2011 to 9.549 million units in 2023, representing a growth of approximately 1137 times. Furthermore, the burgeoning NEV industry has generated new employment opportunities, contributing to economic growth [39].

The intricate link between policy-making institutions, technological advancements, and economic outcomes emphasizes the importance of a comprehensive approach to policy design and implementation. SNA provides valuable insights for understanding these complex relationships and optimizing policy interventions to ensure the continued success of China’s NEV sector [36].

3. Methods and Data

3.1. Data Sources

This study draws data from two primary databases. Market performance data come from the China Association of Automobile Manufacturers (CAAM), covering monthly NEV sales and total vehicle sales from January 2015 to September 2024. Policy network data are extracted from the PKU Law Database, using keywords such as “new energy vehicle”, “electric vehicle”, “fuel cell vehicle”, and “hybrid vehicle” in policy document titles. We collected 279 policy documents each containing comprehensive information, including title, validity period, hierarchical effectiveness, issuing authority, document number, publication date, implementation date, and policy content.

3.2. Network Analysis and Metrics

We employed social network analysis (SNA) to examine NEV policy coordination networks, where government departments serve as nodes and joint policy issuance represents edges. The analysis covers both an overall network (2015–2023) and nine annual networks, using six metrics to evaluate policy coordination:

(1) Network density measures the intensity of policy coordination. A higher density indicates more frequent inter-departmental collaboration in policy-making [40]. For instance, a density of 0.5 suggests that 50% of potential departmental collaborations are actualized through joint policy issuance:

$$D={\displaystyle \frac{2L}{n\left(n-1\right)}}$$

where $L$ represents actual joint policies and $n$ represents departments involved.

(2) Average path length reflects the efficiency of policy coordination. A shorter path length indicates fewer intermediaries needed for policy alignment between departments [41]. For example, a path length of 1.5 suggests most departments can coordinate policies through one or at most two intermediary departments:

$$L={\displaystyle \frac{1}{n\left(n-1\right)}}\sum _{i\ne j}{d}_{ij}$$

where $d_{ij}$ represents the minimum steps needed for policy coordination between departments $i$ and $j$.

(3) Transitivity indicates the stability and maturity of policy coordination [42]. High transitivity suggests that if departments A and B coordinate with C, they likely also coordinate with each other, forming stable policy-making groups:

$$T={\displaystyle \frac{3\times number of triangles}{number of connected triples}}$$

(4) Average clustering coefficient reveals the formation of specialized policy coordination groups [43]. A high coefficient indicates departments tend to form close-knit policy-making clusters, such as groups focusing on technical standards or financial subsidies:

$$C={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{C}_i, C_i={\displaystyle \frac{2L_i}{k_i\left(k_i-1\right)}}$$

where $C_i$ is the clustering coefficient of node $i$, $L_i$ represents the number of edges between node $i$’s neighbors, and $k_i$ is the number of departments connected to department $i$.

(5) Number of components shows the fragmentation of policy coordination [43]. Fewer components suggest more integrated policy-making, while more components indicate relatively independent policy subsystems, such as separate technical and financial policy groups.

(6) Largest component size indicates the scale of core policy coordination [43]. A larger size suggests more departments are integrated into the main policy-making network, enabling comprehensive policy alignment across different aspects of EV development.

These metrics collectively evaluate both the overall structure (through density and components) and specific characteristics (through transitivity and clustering) of policy coordination, providing insights into how different government departments collaborate in EV policy-making.

3.3. Correlation Analysis Using Spearman and Kendall’s Tau

Given the relatively small sample size of our annual network data (n = 9), we employed both Spearman and Kendall’s tau correlation analyses to ensure robust results. These nonparametric methods are particularly suitable for our study as they make no assumptions about data distribution and are resistant to outliers.

The Spearman correlation coefficient (ρ) measures association by ranking variables:

$$\rho =1-{\displaystyle \frac{6{\sum }_{i=1}^n{d}_i^2}{n\left(n^2-1\right)}}$$

where $d_i$ represents the difference between ranks for the ith observation and n is the sample size. For observations with identical values, we used the following:

$$\rho ={\displaystyle \frac{{\sum }_{i=1}^n\left(R\left(X_i\right)-\overline{R_X}\right)\left(R\left(Y_i\right)-\overline{R_Y}\right)}{\sqrt{{\sum }_{i=1}^n{\left(R\left(X_i\right)-\overline{R_X}\right)}^{2{\sum }_{i=1}^n{\left(R\left(Y_i\right)-\overline{R_Y}\right)}^2}}}}$$

where $R\left(X_i\right)$ and $R\left(Y_i\right)$ are the ranked variables [44].

Kendall’s tau (τ) evaluates correlation through concordance of paired observations:

$$\mathsf{\tau }={\displaystyle \frac{2\left(n_c-n_d\right)}{n\left(n-1\right)}}$$

where $n_c$ and $n_d$ represent the numbers of concordant and discordant pairs. For tied observations, we applied the following adjusted formula:

$$\mathsf{\tau }={\displaystyle \frac{n_c-n_d}{\sqrt{\left(n_0-n_1\right)\left(n_0-n_2\right)}}}$$

where $n_0={\displaystyle \frac{n\left(n-1\right)}{2}}$, and $n_1$, $n_2$ are adjustments for ties in X and Y, respectively [44].

Both coefficients range from −1 to 1, with positive values indicating positive correlation and negative values indicating negative correlation. While Spearman’s coefficient offers more intuitive interpretation, Kendall’s tau provides more robust statistical inference for small samples [45]. We report both measures to enhance the reliability of our findings. For statistical significance testing, we used standardized test statistics as both of the coefficients approximate normal distribution even with our sample size.

3.4. The Granger Causality Test

To investigate the causal relationships between policy network characteristics and market performance, we employed the Granger causality test. This methodology, originally proposed by Granger (1969), is particularly suitable for examining temporal predictive relationships in time series data [46]. The fundamental premise is that if variable X Granger causes variable Y, then historical information of X should contribute to predicting future values of Y beyond what could be achieved using only the historical information of Y itself.

The analytical procedure consists of two steps. First, we examined the stationarity of time series using the augmented Dickey–Fuller (ADF) test [47]. For non-stationary series, first-order differencing was applied to achieve stationarity. Second, we conducted the Granger causality test based on a bivariate vector autoregression model:

$$\Delta Y_t={\mathsf{\alpha }}_0+\sum _{i=1}^p{\mathsf{\alpha }}_i\Delta Y_{t-i}+\sum _{j=1}^p{\mathsf{\beta }}_j\Delta X_{t-j}+{\mathsf{\epsilon }}_t$$

where $\Delta Y$ denotes the first difference of market performance indicators (NEV sales or market penetration) at time t, $\Delta X_{t-j}$ represents the first difference of network metrics (Network density, average clustering, or largest component size) with lag j; p is the lag order (set to 2 based on sample size considerations) and ${\mathsf{\epsilon }}_t$ is the error term. The null hypothesis ($H_0:{\mathsf{\beta }}_j=0$ for all j) indicates no Granger causality.

The ADF test model is specified as below:

$$\Delta y_t=\mathsf{\alpha }+\mathsf{\beta }t+\mathsf{\gamma }{y}_{t-1}+\sum _{i=1}^p{\mathsf{\delta }}_i\Delta y_{t-i}+{\mathsf{\epsilon }}_t$$

The F-statistic for testing Granger causality is calculated as follows:

$$F={\displaystyle \frac{\left(RS{S}_R-RS{S}_U\right)/p}{RS{S}_U/\left(T-2p-1\right)}}$$

where $RS{S}_R$ and $RS{S}_U$ are the residual sum of squares from the restricted and unrestricted models, respectively, and T is the sample size.

To capture both immediate and lagged effects, we conducted tests with both one-period and two-period lags. The significance of causal relationships was assessed using F-statistics and corresponding p-values, with significance levels set at 0.05, 0.01, and 0.001. For each pair of variables, bidirectional Granger causality tests were performed to comprehensively understand the causal associations.

3.5. Methodological Innovation and Analysis Tools

This study advances the methodological framework for analyzing NEV policy coordination networks through the integration of structural network analysis, temporal evolution assessment, and market performance evaluation. The analytical approach incorporates sophisticated computational tools to ensure robust and replicable results.

The methodological architecture employs Gephi-0.10.1 as the primary platform for network visualization and metric calculation, enabling dynamic representation of network evolution patterns. Advanced network analyses, correlation studies, and causality testing were conducted through Python, utilizing NetworkX library for complex metrics calculation and specialized packages for statistical analyses.

The innovative aspects of this methodology manifest in several significant dimensions. The integration of temporal and structural analyses represents an advancement beyond conventional static network approaches, enabling the examination of both immediate and long-term effects of policy coordination. The study develops a comprehensive evaluation framework incorporating multiple network metrics—density, path length, transitivity, clustering coefficient, component number, and largest component size. This multi-dimensional approach provides nuanced insights into policy coordination patterns, transcending the limitations of traditional single-metric evaluations.

The application of Granger causality testing to policy network analysis enables rigorous examination of both immediate and lagged effects of network characteristics on market performance, while accounting for potential bidirectional causality. These methodological advances provide quantitative measures for policy coordination concepts and establish robust analytical frameworks for examining causal relationships between network characteristics and market outcomes.

4. Policy Coordination in China’s NEV Industry

4.1. Institutional Distribution of Policy Implementation

Analysis of institutional distribution in the NEV policy documents reveals a hierarchical structure of governmental participation. Among 279 policy documents, the Ministry of Industry and Information Technology (MIIT) demonstrates predominant engagement with 192 documents (35.8%), followed by the State Taxation Administration (STA) and National Development and Reform Commission (NDRC) with 54 (10.1%) and 52 (9.7%) documents, respectively (Figure 1).

The institutional engagement exhibits a clear four-tier structure in policy participation. MIIT occupies the first tier with significantly higher policy involvement than other agencies. The second tier comprises STA, NDRC, and Ministry of Finance (MOF), each participating in 46–54 policies. The third tier includes the Ministry of Science and Technology (MOST), Ministry of Commerce (MOFCOM), and State Administration for Market Regulation (SAMR), contributing to 21–34 policies. The General Administration of Customs (GAC) and National Energy Administration (NEA) form the fourth tier, each participating in approximately 15 policies.

A distinctive feature of the policy framework is the prevalence of joint policy issuance, highlighting the cross-functional coordination in NEV industry development. MIIT serves as the central coordinator, while other agencies provide specialized support in areas such as taxation, development planning, and financial mechanisms. This multi-agency approach has fostered a comprehensive policy support system, addressing technological innovation, market development, and financial support.

The institutional distribution pattern reveals two fundamental characteristics of China’s NEV policy framework: centralized coordination under MIIT leadership and specialized participation from functional departments. This structure enables both strategic planning and tactical implementation across multiple policy dimensions, reflecting the systematic nature and complexity of NEV industry development (Figure 2).

The temporal evolution of China’s NEV policy documents demonstrates three distinct development phases, each characterized by unique patterns of institutional coordination and policy intensity. During the initial phase (2007–2013), policy issuance remained limited, with fewer than five documents annually and predominantly single-agency releases. The fluctuating joint issuance ratios during this period reflected the preliminary nature of coordination mechanisms among government agencies.

The landscape shifted dramatically during the rapid growth phase (2014–2018), marked by a substantial increase in policy output and enhanced inter-agency coordination. This period witnessed unprecedented policy intensity, culminating in 33 documents in 2018, with joint issuances reaching a historic high of 66.67%. This surge in coordinated policy-making signaled a mature recognition of the cross-cutting nature of NEV industry development.

The subsequent stable development phase (2019–2024) has exhibited more balanced policy-making patterns, with annual outputs consistently ranging between 20–30 documents and joint issuance ratios stabilizing at 40–50%. The 2023 data, showing 30 total documents split between 13 joint and 17 single-agency issuances, exemplify this equilibrium between coordinated planning and departmental autonomy in policy implementation.

The longitudinal analysis reveals a clear progression toward more sophisticated policy coordination mechanisms, reflected in both the growing volume of policy documents and the increasing prevalence of joint issuances. This evolution suggests a maturing policy framework that balances the need for comprehensive coordination with the flexibility required for effective sectoral governance in China’s NEV industry development (Figure 3).

China’s NEV policy evolution demonstrates three distinct phases, demarcated by pivotal State Council planning documents. The initial exploration phase (2007–2012) preceded the “Energy-saving and New Energy Vehicle Industry Development Plan (2012–2020)” and exhibited exploratory characteristics. During this period, annual policy output remained below 10 documents, primarily comprising discrete initiatives from NDRC, MIIT, and other agencies, establishing groundwork for subsequent systematic policy development.

The planning-led phase (2012–2020) commenced with the 2012 Development Plan, which established clear industry trajectories and objectives. This phase witnessed substantial policy proliferation, peaking at 74 documents in 2018. MIIT emerged as the dominant policy actor, increasing its annual output from five documents in 2012 to 26 in 2018. Concurrent increases in participation from STA, NDRC, and other agencies fostered a comprehensive multi-agency policy framework.

The strategic upgrade phase (2020–2024), initiated by the “New Energy Vehicle Industry Development Plan (2021–2035)”, positioned NEVs as a strategic emerging industry with development targets extending to 2035. This phase exhibits three key characteristics: diversified institutional participation (exemplified by SAMR’s issuance of seven documents in 2023), comprehensive policy scope (transitioning from financial incentives to multi-dimensional development), and enhanced coordination (maintained joint-issuance ratio of 40–50%).

The State Council planning documents have been instrumental in establishing policy directions, constructing systematic frameworks, and guiding focus from industry cultivation to balanced market development and regulation. This planning-led, coordinated policy approach has effectively supported China’s NEV industry growth while establishing foundations for sustainable development.

4.2. Network Analysis of China’s EV-Related Policies

Quantitative indicators from policy network analysis reveal the following characteristics of institutional collaboration in NEV policy networks.

The network density of 0.199 indicates that, approximately, 19.9% of all possible institutional connections have established policy coordination relationships. This relatively sparse network density ensures necessary policy coordination while avoiding excessive institutional overlap and functional intersection. The average path length (1.26) reflects the mean collaborative distance between policy institutions. Specifically, a path length of 1 indicates direct joint policy issuance between two institutions, while a path length of 2 suggests coordination through an intermediary institution (e.g., MIIT and MOST both issuing joint policies with NDRC but not directly with each other). The average path length of 1.26 demonstrates efficient coordination characteristics in the NEV policy network, as most institutions can achieve policy coordination through direct cooperation or at most one intermediary institution (Figure 4).

The network’s transitivity (0.645) and average clustering coefficient (0.545) are both moderately high, indicating notable institutional clustering within the policy network, with tendencies to form several closely collaborating policy groups. This structural characteristic facilitates specialized coordination mechanisms for different policy issues. The network contains nine components, with the largest component comprising 28 nodes, suggesting the existence of a dominant policy coordination core while maintaining several relatively independent policy sub-networks.

From a policy document perspective, this network structure corresponds highly with the evolution of NEV industry policies. During the early industry cultivation phase, marked by the “Energy-saving and New Energy Vehicle Industry Development Plan (2012–2020)”, policies were mainly led by core departments such as MIIT and NDRC. With the implementation of the “New Energy Vehicle Industry Development Plan (2021–2035)”, the policy network gradually expanded to areas such as market regulation and financial services, forming multiple specialized policy coordination sub-networks. This structure ensures both policy systematization and flexible institutional support for key tasks at different development stages.

To visualize these network characteristics, we employed Gephi-0.10.1 software to construct and analyze the policy institution network, enabling clear representation of institutional relationships and coordination patterns within China’s NEV policy framework (

Table 1. Network analysis metrics of NEV policy coordination.

Metric	Value
Network Density	0.199
Average Path Length	1.261
Transitivity	0.645
Average Clustering Coefficient	0.545
Number of Components	9
Largest Component Size	28

).

4.3. Evolution of Network Analysis Indicators for China’s NEV Policies (2009–2023)

The evolution of China’s NEV policy network from 2009 to 2023 exhibits distinct structural transitions across multiple dimensions. Network density demonstrated an overall upward trajectory, increasing from 0.167 in 2009 to 0.552 in 2022, with a notable peak of 1.0 in 2013 coinciding with the implementation of the “Energy-saving and New Energy Vehicle Industry Development Plan (2012–2020)”. Post-2014, the density stabilized between 0.2–0.4, reflecting the establishment of mature and sustainable coordination mechanisms (Figure 5).

The network’s structural complexity is evidenced by the evolution of the average path length, which increased from 1.0 (2009–2013) to 1.636 in 2018, before moderating to 1.451 in 2023. This trajectory corresponds to the development of multi-level coordination mechanisms, particularly following the 2014 expansion into market regulation and financial services domains. The increased path length, while maintaining operational efficiency, indicates the network’s adaptation to more sophisticated policy requirements.

Network cohesion metrics exhibited highly synchronized patterns. Transitivity increased from 0 in 2009 to 1.0 during 2010–2013, subsequently fluctuated between 0.5–0.8, and ultimately exceeded 0.8 in 2022–2023, indicating significantly enhanced local connectivity. The average clustering coefficient followed a parallel trajectory, reaching its peak of 0.822 in 2022. This synchronized evolution reflects the formation and strengthening of specialized policy coordination clusters, enabling more efficient collaboration within specific policy domains.

The network’s compositional evolution reveals two significant structural trends. First, the number of components reached its maximum of five in 2014–2015, indicating initial diversification of policy coordination groups. Second, the largest component demonstrated steady expansion from two nodes in 2009 to 14 nodes in 2022–2023, reflecting the formation of a robust coordination core. Notably, post-2016, the largest component consistently maintained over 10 nodes, demonstrating enhanced scale effects in policy coordination while preserving necessary functional specialization among departments (

Table 2. Annual network analysis indicators (2014–2023).

Year	Network Density	Average Path Length	Transitivity	Average Clustering Coefficient	Number of Components	Largest Component Size
2014	0.222	1.333	0.667	0.483	5	6
2015	0.205	1.464	0.577	0.499	5	8
2016	0.253	1.582	0.528	0.601	4	11
2017	0.192	1.286	0.778	0.477	7	7
2018	0.229	1.636	0.486	0.656	4	12
2019	0.289	1.536	0.615	0.629	3	8
2020	0.264	1.489	0.623	0.611	4	10
2021	0.379	1.545	0.569	0.744	2	11
2022	0.552	1.363	0.821	0.822	2	14
2023	0.333	1.451	0.797	0.693	4	14

).

The overall evolution of the NEV policy network demonstrates three key characteristics: expanding the scale, optimizing the structure, and strengthening the coordination. The dynamic changes in network indicators reflect a transition from simple early-stage coordination to sophisticated multi-dimensional collaboration mechanisms. This evolution has been particularly pronounced following the implementation of the “New Energy Vehicle Industry Development Plan (2021–2035)” in 2020, as evidenced by improved stability across network indicators. These developments suggest continuous refinement of policy coordination mechanisms, providing increasingly systematic and efficient institutional support for China’s NEV industry development.

5. Correlation Analysis of Policy Network Metrics and NEV Market Penetration

Prior to conducting correlation analysis between policy network metrics and NEV market performance indicators, comprehensive hypothesis testing was performed using four methodologies to determine the appropriate analytical approach. The testing framework incorporated the Shapiro–Wilk normality test, Skewness and Kurtosis analysis, outlier detection, and the Durbin–Watson autocorrelation test. The Shapiro–Wilk test assesses data normality (p > 0.05 indicating normal distribution); Skewness measures distribution symmetry (positive values indicating right skew, negative values indicating left skew); Kurtosis evaluates distribution peakedness (positive values indicating leptokurtic distribution, negative values indicating platykurtic distribution); and the Durbin–Watson test examines time series autocorrelation (values approaching 2 indicating absence of autocorrelation) (

Table 3. Hypothesis testing results for network and market performance indicators.

Indicator	Shapiro–Wilk Statistic	p-Value	Skewness	Kurtosis	Number of Outliers	DW Statistic
Network Density	0.852	0.079	1.332	0.904	1	0.909
Average Path Length	0.969	0.886	−0.477	−0.591	1	0.136
Transitivity	0.89	0.202	0.373	−1.372	0	0.341
Average Clustering Coefficient	0.97	0.898	0.099	−0.713	0	0.238
Number of Components	0.896	0.231	0.633	0.06	1	0.39
Largest Component Size	0.926	0.445	0.051	−1.247	0	0.458
NEV sales	0.76	0.007	1.229	−0.026	1	0.526
Market Penetration	0.769	0.009	1.106	−0.402	1	0.539

).

The test results reveal that network metrics (network density, average path length, transitivity, average clustering coefficient, number of components, and largest component size) demonstrate normal distribution (p > 0.05), while market performance indicators (NEV sales and market penetration) exhibit non-normal distribution (p = 0.007 and 0.009, respectively). Most indicators show significant time series autocorrelation, evidenced by Durbin–Watson statistics substantially below 2. Furthermore, network density, average path length, number of components, NEV sales, and market penetration contain outliers.

Given these findings—non-normal distribution in key variables, presence of outliers, limited sample size (n = 9), and significant time series autocorrelation—the Spearman rank correlation coefficient was selected as the primary analytical method, supplemented by Kendall’s tau for validation. This methodological choice is justified by Spearman correlation’s non-parametric nature, robustness to outliers, and suitability for small sample analysis, aligning with the dataset’s characteristics.

The correlation analysis reveals complex relationships between policy network characteristics and NEV market performance, supported by both statistical evidence and visual analysis (

Table 4. Correlation analysis of network metrics and market performance indicators.

Network Metric	Market Metric	Spearman rho	Spearman p	Kendall Tau	Kendall p	Significance
Network Density	NEV sales	0.8000	0.0096	0.6111	0.0247	***
Network Density	Market Penetration	0.8500	0.0037	0.6667	0.0127	***
Average Path Length	NEV sales	−0.1833	0.6368	−0.1111	0.7614
Average Path Length	Market Penetration	−0.2333	0.5457	−0.1667	0.6122
Transitivity	NEV sales	0.5000	0.1705	0.3333	0.2595
Transitivity	Market Penetration	0.5667	0.1116	0.3889	0.1802
Average Clustering Coefficient	NEV sales	0.8500	0.0037	0.6667	0.0127	***
Average Clustering Coefficient	Market Penetration	0.8333	0.0053	0.6111	0.0247	***
Number of Components	NEV sales	−0.6033	0.0854	−0.4642	0.0983	*
Number of Components	Market Penetration	−0.6470	0.0596	−0.5261	0.0610	*
Largest Component Size	NEV sales	0.7342	0.0243	0.6093	0.0260	**
Largest Component Size	Market Penetration	0.6583	0.0539	0.5512	0.0440	*

Note: Significance levels: *** p < 0.01; ** p < 0.05; * p < 0.1.

).

The correlation analysis results demonstrate several significant relationships between network characteristics and market performance indicators. Network density exhibits particularly strong positive correlations with both NEV sales (ρ = 0.800, p < 0.01) and market penetration (ρ = 0.850, p < 0.01), suggesting that enhanced policy network cohesion strongly corresponds with market development. This finding indicates that denser policy coordination networks, characterized by more frequent and comprehensive institutional interactions, may facilitate more effective market development.

The average clustering coefficient shows similarly robust positive correlations with NEV sales (ρ = 0.850, p < 0.01) and market penetration (ρ = 0.833, p < 0.01). These strong correlations suggest that localized policy coordination clusters play a crucial role in market development, potentially by enabling more efficient policy implementation and coordination within specialized policy domains.

The largest component size demonstrates significant positive correlation with NEV sales (ρ = 0.734, p < 0.05) and marginally significant correlation with market penetration (ρ = 0.658, p < 0.1). These correlations suggest that the scale of core policy coordination networks positively influences market outcomes, potentially through enhanced policy coherence and implementation efficiency within the main coordination group.

Conversely, the number of components shows marginally significant negative correlations with both market indicators (NEV sales: ρ = −0.603, p < 0.1; market penetration: ρ = −0.647, p < 0.1). This negative relationship suggests that policy network fragmentation may impede market development, possibly by increasing coordination costs and reducing policy coherence.

Notably, average path length shows weak and insignificant correlations with market performance indicators (NEV sales: ρ = −0.183, p > 0.1; market penetration: ρ = −0.233, p > 0.1), suggesting that the efficiency of indirect policy coordination may be less critical than direct coordination strength (Figure 6).

The temporal analysis reveals particularly interesting patterns in the co-evolution of network characteristics and market performance. Post-2021, the data show a marked acceleration in market performance metrics coinciding with significant changes in network structure. During this period, network density and average clustering coefficient demonstrated upward trends, while the number of components decreased, suggesting network optimization. This synchronized temporal variation provides additional support for the correlation analysis results and indicates that optimized policy network structures (characterized by higher density, enhanced clustering, and reduced fragmentation) correspond with improved market outcomes.

These findings collectively suggest that policy network optimization, particularly through enhanced coordination density and clustering, may contribute significantly to NEV market development. The negative impact of network fragmentation and the positive influence of core component size further emphasize the importance of maintaining cohesive and well-structured policy coordination networks.

6. Granger Causality Analysis

Following the correlation analysis, we conducted Granger causality tests to examine the temporal causal relationships between policy network characteristics and market performance indicators. The test results reveal complex bidirectional causal relationships with distinct temporal patterns and varying intensities (

Table 5. Bidirectional Granger causality test results between network characteristics and market performance.

Variables	Lag	F-Statistic	p-Value
Network Density→NEV sales	1	0.003	0.9551
NEV sales→Network Density	1	0.723	0.395
Network Density→NEV sales	2	284.051	0
NEV sales→Network Density	2	27.952	0
Network Density→Market Penetration	1	0.006	0.9385
Market Penetration→Network Density	1	0.702	0.402
Network Density→Market Penetration	2	95.524	0
Market Penetration→Network Density	2	16.857	0.0002
Average Clustering Coefficient→NEV sales	1	0.151	0.698
NEV sales→Average Clustering Coefficient	1	0.22	0.6393
Average Clustering Coefficient→NEV sales	2	579.087	0
NEV sales→Average Clustering Coefficient	2	1.645	0.4394
Average Clustering Coefficient→Market Penetration	1	0.161	0.6885
Market Penetration→Average Clustering Coefficient	1	0.033	0.8551
Average Clustering Coefficient→Market Penetration	2	281,486.748	0
Market Penetration→Average Clustering Coefficient	2	1.342	0.5111
Largest Component Size→NEV sales	1	4.152	0.0416
NEV sales→Largest Component Size	1	0.018	0.8924
Largest Component Size→NEV sales	2	1522.706	0

).

Policy network characteristics demonstrate pronounced time-lagged effects on market performance. The largest component size shows immediate causality on NEV sales in the short term (one-period lag) (F = 4.152, p < 0.05), with this effect strengthening substantially over two periods (F = 1522.706, p < 0.001). In contrast, the network density and average clustering coefficient exhibit no significant short-term causality (p > 0.05) but demonstrate strong long-term effects. Specifically, the average clustering coefficient shows the most substantial long-term impact on market penetration (F = 281,486.748, p < 0.001), while the network density significantly influences both NEV sales (F = 284.051, p < 0.001) and market penetration (F = 95.524, p < 0.001).

The reverse causality analysis reveals more nuanced patterns. Market performance indicators only demonstrate significant long-term effects on the network density. Both NEV sales (F = 27.952, p < 0.001) and market penetration (F = 16.857, p < 0.001) show significant causal effects on the network density with two-period lags. Notably, market performance indicators show no significant causal influence on either the average clustering coefficient or the largest component size, regardless of the time frame (p > 0.05).

These findings extend our correlation analysis by revealing underlying causal mechanisms. First, there exists an asymmetric causal relationship where policy network characteristics predominantly drive market development. Second, the largest component size, reflecting the core of policy coordination, demonstrates the most immediate and persistent influence on market performance. Finally, the market performance feedback primarily manifests in its impact on the network density, suggesting that market development promotes overall policy network cohesion without significantly altering local network structural characteristics.

7. Conclusions and Policy Recommendations

This study employed social network analysis and Granger causality tests to systematically examine the evolution characteristics of China’s NEV policy networks from 2015–2023 and their impact on market performance. The research reveals both correlative and causal relationships between policy coordination and NEV market development. Network density shows highly significant positive correlations with NEV sales and market penetration (ρ = 0.800 and 0.850, p < 0.01), with strong long-term causal effects (F = 284.051 and 95.524, p < 0.001). Average clustering coefficient demonstrates highly significant positive correlations with market performance indicators (ρ = 0.850 and 0.833, p < 0.01), exhibiting particularly strong long-term causal influence on market penetration (F = 281,486.748, p < 0.001). Additionally, largest component size shows both significant positive correlation (ρ = 0.734, p < 0.05) and immediate causal effects (F = 4.152, p < 0.05) on NEV sales, with strengthened long-term impact (F = 1522.706, p < 0.001).

From the perspective of policy network evolution, NEV policy coordination from 2015–2023 shows distinct phase characteristics. The network density showed an overall upward trend, rising from 0.205 in 2015 to 0.552 in 2022; the average clustering coefficient gradually increased, reaching a peak of 0.822 in 2022; the largest component size continuously expanded from eight nodes in 2015 to 14 nodes in 2023. These evolutionary characteristics reflect the continuous optimization of policy coordination mechanisms, providing strong support for rapid NEV market development. Notably, the Granger causality analysis reveals that market development also exerts significant feedback effects on network density in the long term (F = 27.952 and 16.857, p < 0.001), indicating a virtuous cycle between policy coordination and market growth.

Based on the research results, this paper proposes the following policy recommendations: First, establish a multi-level policy coordination system, creating a coordination mechanism led by NDRC with participation from MIIT, MOF, and other departments, improving central–local policy linkage. Second, optimize policy network structure, particularly strengthening the core component’s role given its immediate market impact, while clarifying departmental responsibilities and establishing policy information sharing platforms. Third, establish dynamic policy adjustment mechanisms, considering the time-lagged effects revealed by causality analysis, to ensure timely and effective policy responses. Fourth, strengthen coordinated policy implementation supervision, promoting supervisory information sharing and establishing cross-department joint supervision mechanisms. Finally, improve supporting institutional construction, enhancing legal regulations and standards systems, strengthening talent cultivation and technological innovation.

While this study provides valuable insights into NEV policy coordination networks, several limitations should be acknowledged. The current analysis focuses primarily on national-level policies, which may not fully capture the complexity of policy implementation across different administrative levels. The network metrics, though quantitative, might not completely reflect the nuanced coordination patterns between central and local governments. Additionally, the relatively short time span of the dataset (2015–2023) may limit our ability to fully capture long-term policy effects and market development patterns.

These limitations suggest promising directions for future research. First, extending the analysis to provincial policy level would enable a more comprehensive understanding of central–local policy coordination effects. Second, introducing international comparative perspectives would enrich our understanding of policy coordination mechanisms in different institutional contexts. As the NEV industry continues to develop, optimizing policy coordination mechanisms will provide stronger institutional support for high-quality industry development.