Analysis of Technological Innovation Efficiency in Listed New Energy Vehicle Enterprises Under the Carbon Neutrality Framework Based on Two-Stage Dynamic Network DEA and a GRA Model

Abstract

Technological innovation and the efficiency of resource allocation in Chinese new energy vehicle enterprises represent critical factors influencing the sustainable development of the industry. By applying a two-stage dynamic network DEA model to analyze the comprehensive and stage-specific technological innovation efficiency of 13 A-share-listed new energy vehicle enterprises between 2017 and 2024, this study reveals that both overall and phase-specific innovation efficiencies remain below optimal levels. Moreover, the average technological R&D efficiency across these firms is found to be lower than their average achievement transformation efficiency, highlighting the urgent need to improve innovation performance in this sector. Grey relational analysis of influencing factors identifies six key determinants of technological innovation efficiency: the shareholding ratio of the largest shareholder, R&D investment intensity, the proportion of employees holding bachelor’s degrees or higher, management capability, return on equity, and total asset turnover. In comparison, government subsidies and total assets exhibit relatively limited influence on technological innovation efficiency.

1. Introduction

In 2024, automobile production and sales in China reached 31.282 million and 31.436 million units, reflecting year-on-year increases of 3.7% and 4.5%, respectively. Notably, new energy vehicle (NEV) production and sales amounted to 12.888 million and 12.866 million units, registering significant year-on-year growth of 34.4% and 35.5%, respectively. New energy vehicles, characterized by advanced technology, low carbon emissions, and high energy efficiency, have become a key priority in national development strategies worldwide. Developed countries such as the United States, Japan, and Germany have introduced policies related to “Carbon Neutrality” and invested significant human, material, and financial resources to support the development of new energy vehicles. China has also introduced multiple nationally led policies such as the “Torch Plan”, “863 Project”, and “975 Project”, which clearly highlight the need to accelerate the cultivation and promotion of energy-saving and new energy vehicle industry developments, implement subsidies for new energy vehicles, and promote these vehicles in cities. China is continuously adjusting its new energy industry policies in line with industrial development trends, implementing a comprehensive set of supportive measures encompassing macro-level coordination, fiscal subsidies, tax incentives, financial services, technological research and development, and infrastructure construction [1,2]. The new energy vehicle (NEV) industry has emerged as a pivotal driver of China’s economic structural transformation and upgrading, facilitating the transition to a low-carbon, sustainable, and high-quality development model. Research and innovation efficiency in the NEV sector have attracted considerable attention from academics and elsewhere. In recent years, guided by the goal of carbon neutrality, the strategic use of policy instruments to optimize the allocation of innovation resources and enhance technological innovation efficiency in the NEV industry has become a key focus of scholarly inquiry [3].

2. Materials and Methods

The methodology for measuring efficiency has undergone a significant transformation, evolving from parametric frontier approaches to advanced non-parametric network models. Within the parametric framework, Stochastic Frontier Analysis (SFA), introduced by Aigner et al. (1977) [4], has become the predominant methodological approach. Recent advancements have further strengthened its utility through integration with Bayesian estimation techniques and extensions to panel data specifications. Among non-parametric methods, Data Envelopment Analysis (DEA) offers greater flexibility and has progressed substantially—from the original CCR model (Charnes et al., 1978) [5] and BCC model (Banker et al., 1984) [6]—to include directional distance functions (Chung et al., 1997) [7], Bootstrap DEA (Simar & Wilson, 1998, 2007) [8,9], and meta-frontier analysis (Battese & Rao, 2002) [10]. The DEA framework has also shifted from a “static black box” representation to a “dynamic network” paradigm, driven by methodological innovations such as the two-stage network DEA model (Färe & Grosskopf, 2000) [11], the dynamic SBM model (Tone & Tsutsui, 2010) [12], and the non-radial SBM model (Tone, 2001) [13]. Alongside these developments, the conceptual understanding of innovation measurement has broadened beyond narrow technical innovation to incorporate multidimensional forms, including organizational and marketing innovations, and is increasingly aligned with sustainable development objectives. This evolution is clearly reflected in the integration of Green Total Factor Productivity (GTFP) and circular economy principles into performance evaluation frameworks, marking the field’s transition toward a more comprehensive paradigm that balances economic, environmental, and social dimensions.

Literature analysis reveals that scholars both domestically and abroad predominantly employ the entropy method, stochastic frontier analysis (SFA), and data envelopment analysis (DEA) to assess the technological innovation efficiency of new energy vehicles [14]. Among them, the combination of data envelopment analysis and the DEA with the Malmquist index method is most widely used [15,16]. To date, no scholar has employed a two-stage dynamic network DEA model to assess the technological innovation efficiency of listed new energy vehicle companies, nor has any scholar conducted a comprehensive analysis of the influencing factors using GRA. In the past 10 years, both national policies and marketization have vigorously promoted new energy vehicles. Using a two-stage dynamic network DEA model and GRA to study the comprehensive technological innovation efficiency and influencing factors of Chinese new energy vehicles not only assists practical development but also provides a theoretical basis for the government to formulate new energy vehicle industry policies [17,18].

2.1. Dynamic Network DEA Model

Tone and Tsutsui (2014) [19] developed a dynamic network DEA model by integrating the slack-based network DEA model with the slack-based dynamic DEA model, establishing a DEA framework that accounts for both network structure and inter-period dependencies [20,21]. In the horizontal dimension, dynamic network DEA connects two departments through the network structure in each period; in the vertical dimension, it combines the network structure through carry-over methods. This model enables the assessment of comprehensive efficiency, period-specific dynamic change efficiency, and department-specific dynamic change efficiency over the entire observation period.

The basic model of the two-stage dynamic network DEA model is shown in Figure 1 below.

Assuming the evaluation period consists of T time periods, the production process of a decision-making unit in any given period can be divided into two distinct sub-stages. In the t-th period (t = 1, …, T), the decision-making unit employs m inputs $x_{ij}^t(i=1,\dots ,m;j=1,\dots ,n)$ in the first stage to generate d intermediate outputs $z_{qj}^t(q=1,\dots ,d;j=1,\dots ,n)$, which are subsequently used in the second stage to produce r final outputs $y_{rj}^t(r=1,\dots ,s;j=1,\dots ,n)$. In the first stage, the transition between period t and period t + 1 is facilitated by carryover variables $c_{p_{1}j}^t(p_1=1,\dots ,c_1;j=1,\dots ,n)$; likewise, in the second stage, the linkage between these consecutive periods is established through the same carryover variables $c_{p_{2}j}^t(p_2=1,\dots ,c_2;j=1,\dots ,n)$. Assuming that all carry-over variables are desirable and that the intermediate variables between the first and second stages in period t are freely disposable, the dynamic network SBM model for evaluating the overall efficiency of the k-th decision-making unit over the evaluation period T can be expressed as Equation (1):

$$E_K^T=\min {\displaystyle \frac{{\displaystyle \sum _{i=1}^T{W}_t\left\{w_1\left(1-\frac{1}{m}{\displaystyle \sum _{i=1}^m\frac{s_i^{t-}}{x_{ik}^t}}\right)+w_2\left(1-\frac{1}{d}{\displaystyle \sum _{q=1}^d\frac{s_q^{t-}}{z_{qk}^t}}\right)\right\}}}{{\displaystyle \sum _{i=1}^T{W}_t\left\{w_1\left[1+\frac{1}{d+c_1}\left({\displaystyle \sum _{q=1}^d\frac{s_q^{t+}}{z_{qk}^t}+{\displaystyle \sum _{p_1=1}^{c_1}\frac{s_{p_1}^{t+}}{c_{p_{1}k}^t}}}\right)\right]+w_2\left[1+\frac{1}{s+c_2}\left({\displaystyle \sum _{r=1}^s\frac{s_r^{t+}}{y_{rk}^t}+{\displaystyle \sum _{p_2=1}^{c_2}\frac{s_{p_2}^{t+}}{c_{p_{2}j}^t}}}\right)\right]\right\}}}}$$

(1)

The constraint conditions are as follows:

$${\displaystyle \sum _{j=1}^n{\lambda }_j^t}{x}_{ij}^t+s_i^{t-}=x_{ik}^t(\forall i,t)$$

$${\displaystyle \sum _{j=1}^n{\lambda }_j^t}{z}_{qj}^t-s_q^{t+}=z_{qk}^t(\forall q,t)$$

$${\displaystyle \sum _{j=1}^n{\lambda }_j^t}{c}_{p_{1}j}^t-s_{p_1}^{t+}=c_{p_{1}k}^t(\forall t)$$

$${\displaystyle \sum _{j=1}^n{\lambda }_j^t}{c}_{p_{1}j}^t={\displaystyle \sum _{j=1}^n{\lambda }_j^{t+1}}{c}_{p_{1}j}^t$$

$${\displaystyle \sum _{j=1}^n{\eta }_j^t}{z}_{qj}^t+s_q^{t-}=z_{qk}^t(\forall q,t)$$

$${\displaystyle \sum _{j=1}^n{\eta }_j^t}{c}_{p_{2}j}^t-s_{p_2}^{t+}=c_{p_{2}k}^t(\forall t)$$

$${\displaystyle \sum _{j=1}^n{\eta }_j^t}{c}_{p_{2}j}^t={\displaystyle \sum _{j=1}^n{\eta }_j^{t+1}}{c}_{p_{2}j}^t$$

$${\displaystyle \sum _{j=1}^n{\eta }_j^t}{y}_{rj}^t-s_r^{t+}=y_{rk}^t(\forall r,t)$$

$${\displaystyle \sum _{j=1}^n{\lambda }_j^t}=1(\forall t)$$

$${\lambda }_j^t,{\eta }_j^t,s_i^{t-},s_q^{t+},s_{p_1}^{t+},s_{p_2}^{t+},s_q^{t-},s_r^{t+}\ge 0,\forall i,q,r,j,t$$

In this context, ${\lambda }_j^t,{\eta }_j^t$ represent the input and output variables of the first stage and the second stage in period t, respectively, while $s_i^{t-},s_q^{t+},s_{p_1}^{t+},s_{p_2}^{t+},s_q^{t-},s_r^{t+}$ denote the slack variables associated with each evaluation indicator; $w_1,w_2$ denote the weights assigned to the efficiencies of the first stage and the second stage, respectively, in the overall efficiency calculation of the decision-making unit for a single period, and $w_1+w_2=1$; $W_t$ represents the weight of the overall efficiency of the decision-making unit in period t relative to the total efficiency over the evaluation period t and ${\displaystyle \sum _{t=1}^T{W}_t}=1$

2.2. Grey Relational Analysis

Grey Relational Analysis Method: The relationship between sub-factors and their parent factors is indicated by the correlation magnitude. A high correlation suggests a significant influence of sub-factors on the parent factors, whereas a low correlation implies a relatively minor influence.

The grey relational analysis method boasts advantages such as low data requirements and the capacity to intuitively assess the extent of influence exerted by sub-factors on the parent factor. In this study, the influence degree of the following eight sub-factors is analyzed by taking the comprehensive technological innovation efficiency value of listed new energy vehicle companies as the parent factor: the shareholding ratio of the largest shareholder, R&D investment, the proportion of employees with a bachelor’s degree or higher, management capability, return on net assets, total asset turnover rate, government subsidies, and total assets [22,23,24]. The sub-factors are then ranked based on the magnitude of their correlation degree, and an advantage analysis is performed.

3. Results

3.1. Data Sources

We selected new energy vehicle enterprises listed on the A-share markets of the Shanghai Stock Exchange and Shenzhen Stock Exchange from 2017 to 2024 as research samples. ST- and *ST-listed enterprises were excluded. The two-stage dynamic network DEA model requires no negative output; thus, enterprises with negative net profit during the 8 years from 2017 to 2024 were excluded. Finally, 13 listed new energy vehicle enterprises from 2017 to 2024 were selected as research objects. The sample data were derived from the annual reports of the enterprises. For instances of missing values, supplementary information was collected from Baidu and corporate financial reports. To construct a complete time series, linear interpolation was applied to impute missing observations. To ensure the reliability of the data processing procedure, the following validation protocols were implemented: first, an internal consistency check was conducted using random simulation-based imputation; second, a sensitivity analysis was carried out by comparing results derived from alternative interpolation methods, including spline interpolation; third, descriptive statistics were examined before and after imputation. The results demonstrate that the imputation process did not introduce significant bias into the data distribution, thus supporting the robustness of the subsequent findings. The sample selection is shown in

Table 1. Basic information of 13 listed new energy vehicle companies.

Listed Company	Stock Code	A-Share?	Registered Company Address	Nature of Controlling Shareholders
Aote Xun	002227SZ	A-share	Guangdong Province	Other
BYD	002594SZ	A-share	Guangdong Province	Private enterprise
Dongfeng Technology	600081SH	A-share	Shanghai	State-owned enterprise
Dongfeng Motor	600006SH	A-share	Hubei Province	State-owned enterprise
GAC Group	601238SH	A-share	Guangdong Province	State-owned enterprise
Ningbo Yunsheng	600366SH	A-share	Zhejiang Province	Other
SAIC Motor	600104SH	A-share	Shanghai	Other
Universal Money Tide	000559SZ	A-share	Zhejiang Province	Private enterprise
Wolong Electric	600580SH	A-share	Zhejiang Province	Other
Yutong Bus	600066SH	A-share	Henan Province	Private enterprise
Changan Automobile	000625SZ	A-share	Chongqing	State-owned enterprise
Great Wall Motor	601633SH	A-share	Hebei Province	Other
Zhongtong Bus	000957SZ	A-share	Shandong Province	Other

.

3.2. Indicator Selection and Sample Data Analysis

The two-stage dynamic network DEA model was used to evaluate the technological innovation and technical efficiencies of listed new energy vehicle companies in two stages: the technology research and development stage and the achievement transformation stage. There were nine input and output indicators, including fixed assets, R&D investment, number of enterprise R&D personnel, number of patent authorizations, technology asset ratio, number of valid invention patents, total number of employees, and business income of the enterprise [25,26].

Fixed assets measures the capital invested by enterprises and plays a role in the production process over the long term, while R&D investment refers to the expenses used for product research and development and the number of enterprise R&D personnel can reflect the true state of the enterprise’s technological innovation efficiency. The patent authorization number is a general indicator to measure the level of knowledge output in innovation activities, and the technology asset ratio reflects the efficiency of the enterprise’s intangible asset research and development results. The effective invention patent number refers to the number of invention patents within the validity period; and the total number of employees refers to the labor force invested by enterprises, while the main business income refers to the income obtained from engaging in main business and net profit refers to after-tax profit [27,28].

The research model is shown in Figure 2. Based on the research model, nine two-stage dynamic network DEA input–output indicators (

Table 2. Input, output, connection, and carry-over indicators.

	Technical Research and Development Stage		Intermediate Indicator	Achievement Transformation Stage		Carry-Over Indicator
	Input Indicators	Output Indicators	The First Stage/The Second Stage	Input Indicators	Output Indicators	T Period– T + 1 Period
DNSBM	1. Fixed assets 2. R&D investment 3. Number of R&D personnel in the enterprise	6. Technical asset ratio	5. Number of patents granted	5. Number of patents granted 7. Total number of employees	8. Main business income of the enterprise	The first stage—4. Patent for invention (carry-over) The second stage—9. Net Profit

) were constructed for the 13 listed new energy vehicle companies. The descriptive statistics of these indicators are shown in

Table 3. Descriptive statistics of input and output indicators of 13 listed companies.

	Quantity	Min	Max	Average	STDEV
Fixed Assets	104	12,070,845.400	83,056,007,151.500	11,376,251,463.406	16,095,434,153.687
R&D Investment	104	23,786,500.000	15,385,012,641.180	1,945,638,282.373	2,934,162,118.755
Number of R&D Personnel in the Enterprise	104	174.000	35,788.000	5497.817	7545.075
Number of Valid Invention Patents	104	1.000	1368.000	189.471	313.337
Number of Patents Granted	104	2.000	7533.000	1134.365	1479.375
Technical Asset Ratio	104	0.008	0.101	0.039	0.021
Total Number of Employees	104	503.000	229,154.000	33,018.390	52,338.917
Main Business Income of the Enterprise	104	251,755,817.510	887,626,207,288.410	80,999,114,958.803	188,041,969,449.538
Net Profit	104	8,779,746.740	48,404,663,401.860	5,077,189,744.771	10,692,756,958.145
Significant Figure	104

.

4. Discussion

4.1. Data Processing

Using MaxDEA X v12.2, we calculated the comprehensive technological innovation efficiency θ, technological research and development stage efficiency θ₁, and achievement transformation stage efficiency θ₂ of the 13 sample enterprises from 2017 to 2024 using a two-stage dynamic network DEA model.

Table 4. Comprehensive technological innovation efficiency and two-stage efficiency scores of 13 new energy vehicle listed companies.

Listed Companies	θ	Rank	θ1	Rank	θ2	Rank
Aote Xun	1	1	1	1	1	1
BYD	1	1	1	1	1	1
Dongfeng Technology	0.688	7	0.610	9	0.629	9
Dongfeng Motor	0.831	5	1	1	0.939	6
GAC Group	0.422	10	0.776	8	1	1
Ningbo Yunsheng	0.933	3	0.383	11	0.638	8
SAIC Motor	0.692	6	0.959	5	0.602	11
Universal Money Tide	0.267	11	0.156	13	0.310	13
Wolong Electric	0.629	8	0.528	10	0.608	10
Yutong Bus	0.110	13	0.259	12	0.343	12
Changan Automobile	0.456	9	0.974	4	0.690	7
Great Wall Motor	0.154	12	0.844	7	0.972	5
Zhongtong Bus	0.933	4	0.896	6	1	1
Average	0.624		0.722		0.749

Note: θ represents the comprehensive technological innovation efficiency, θ₁ represents the efficiency of the technology research and development stage, and θ₂ represents the efficiency of the achievement transformation stage.

presents the results of these calculations.

Table 5. Effective proportion of technological innovation efficiency of 13 listed new energy vehicle companies in China.

	Overall Efficiency		The First Stage		The Second Stage
	Effective Quantity	Proportion	Effective Quantity	Proportion	Effective Quantity	Proportion
Effective	2	15.38%	3	23.08%	4	30.77%
Invalid	11	84.62%	10	76.92%	9	69.23%

shows that the proportion of the 13 listed new energy vehicle companies with effective comprehensive technological innovation efficiency is 15.38%, and the proportion of those with effective technological research and development stage efficiency is 23.08%.

In the overall stage, the comprehensive technological innovation efficiency score (θ) ranges from 0.110 to 1.000, with an average of 0.624. Only two companies—Aote Xun and BYD—achieve effective technological research and development efficiency values of 1, accounting for 15.38% of the total. Among the remaining 11 inefficient companies, 4 have efficiency scores lower than the average value of 0.6242.

In the first stage, the efficiency score of the technology research and development stage (θ₁) ranges from 0.156 to 1.000, with an average of 0.722. Only three companies—Aote Xun, BYD, and Dongfeng Motor—have a technology research and development efficiency value of 1, achieving technical effectiveness, accounting for 23.08% of the total. Among the remaining ten companies with low efficiency, four have efficiency scores lower than the average value of 0.7219.

In the second stage, the score of the efficiency of achievement transformation (θ₂) ranges from 0.310 to 1.000, with an average value of 0.749. Among them, four enterprises—Aote Xun, BYD, Guangzhou Automobile Group, and Zhongtong Bus—achieved effective results, with an efficiency score of 1, accounting for 30.77% of the total. Among the remaining nine inefficient enterprises, six have efficiency scores lower than the average value of 0.7485.

The average comprehensive technological innovation efficiency of the 13 listed new energy vehicle companies is not high, and the average efficiency value in the first stage of technological research and development is lower than that in the second stage of achievement transformation. Only two companies (Aote Xun and BYD) achieved effective values in comprehensive technological innovation efficiency, technological research and development stage efficiency, and achievement transformation stage efficiency during the analyzed period. The remaining 11 companies that performed well in technological research and development stage efficiency did not perform well in the achievement transformation stage, and vice versa. These companies can improve their overall comprehensive technological innovation efficiency by improving their technological research and development efficiency, achievement transformation efficiency, or both.

Table 6. Technological innovation efficiency of listed new energy vehicle enterprises with different properties.

Score	State-Owned Enterprises (4)	Private Enterprises (3)	Others (6)	Kruskal–Wallis H	p-Value	Effect Size (η2)	Statistical Conclusions
θ	0.599	0.459	0.724	4.740	0.093	0.365	No Significant Difference
θ1	0.840	0.472	0.768	6.200	0.045 *	0.477	Significant Difference
θ2	0.815	0.551	0.803	5.640	0.060	0.434	Marginally Significant

Note: ***, **, and * represent significant values at the 1%, 5%, and 10% levels, respectively.

presents a statistical analysis of efficiency based on the nature of the sample enterprises. From the perspective of property rights of sample enterprises, state-owned enterprises exhibit higher comprehensive technological innovation efficiency (0.599), technological research and development stage efficiency (0.840), and commercial transformation stage efficiency (0.815) compared to private enterprises. The superior development quality of state-owned enterprises in the new energy vehicle sector compared to private enterprises can be attributed to the policy support, locational advantages, and natural endowments enjoyed by state-owned enterprises.

However, further statistical significance testing (Kruskal–Wallis H test) provides a more precise interpretation of these findings. Significant differences were observed in the efficiency of the technology R&D phase (θ₁). State-owned enterprises (0.840) demonstrated significantly higher efficiency than private enterprises (0.472) (p = 0.045), strongly confirming the efficiency advantage of state-owned enterprises in this phase. This is closely related to the fact that state-owned enterprises typically receive stronger policy support, R&D subsidies, and possess inherent advantages in location and natural endowments.

In the commercial transformation phase (θ₂) and comprehensive efficiency (θ), the inter-group differences did not reach strict statistical significance levels (p-values of 0.060 and 0.093, respectively), but the substantial effect sizes suggest that the differences remain meaningful. Notably, “other enterprises” exhibited the best performance across all efficiency metrics, indicating that ownership type is not the sole determinant of efficiency. This may include enterprises with superior management mechanisms, market flexibility, or technological approaches.

The study confirms significant efficiency differences among enterprises of different ownership types, particularly in the technology R&D phase, with the advantage of state-owned enterprises being statistically established in this stage. However, caution is warranted when generalizing the conclusion that “state-owned enterprises are superior to private enterprises” across the entire innovation process. Future analyses should focus more on the underlying causes of the high efficiency of “other enterprises” and incorporate covariates such as enterprise scale and R&D investment intensity for a more in-depth exploration.

4.2. Grey Correlation Analysis of Factors Affecting Technological Innovation Efficiency

The analysis results of the two-stage dynamic network DEA model indicate that the comprehensive technological innovation efficiency of currently listed Chinese new energy vehicle companies is low. Improving the technological innovation efficiency of listed new energy vehicle companies is an urgent problem to be solved. The factors affecting the technological innovation efficiency of these companies include internal factors and external economic and environmental factors. In order to further explore the deep-seated influencing factors behind this efficiency, the grey relational analysis method (GRA) was used to evaluate the degree of correlation between sub-factors and parent factors in the system. The technological innovation efficiency was selected as the parent factor in the two-stage dynamic network DEA model, and eight sub-factors—total assets, shareholding ratio of the largest shareholder, R&D investment, proportion of employees with bachelor’s degree or above, government subsidies, management ability, return on net assets, and total asset turnover rate—were selected for grey relational analysis.

Total assets (X₁) refers to the total assets of the enterprise at the end of the year. The shareholding ratio of the largest shareholder (X₂), which holds a relatively high proportion of shares, is more deeply tied to the enterprise’s operating performance, with a higher degree of interest correlation, and is more likely to determine the enterprise’s operational decisions and development strategies. R&D investment (X₃) plays an important role in the innovation process of the enterprise; an increase in R&D investment improves the enterprise’s technological level and product quality. The proportion of employees with a bachelor’s degree or above (X₄) is the ratio of employees with a bachelor’s degree or above to the total number of employees. Government subsidies (X₅), which enterprises can obtain through the sale of qualified and complete new energy vehicles, can provide corresponding sales subsidies afterwards; when enterprises carry out related research and development activities, they can also obtain corresponding research and development subsidies in this process. Management ability (X₆) and efficiency help enterprises carry out technological innovation activities. Return on net assets (X₇) is composed of the ratio of net profit to total assets, while the total asset turnover rate (X₈) is composed of the ratio of net operating income to average total assets.

The indicators, along with their definitions and formulae, are shown in

Table 7. Factors affecting the technological innovation efficiency of listed new energy vehicle companies.

Variable	Formula	Symbol
Comprehensive Technological Innovation Efficiency	Comprehensive technological innovation efficiency	θ
Total Assets	Total assets (CNY)	X1
Shareholding Ratio of the Largest Shareholder	Shareholding ratio of the largest shareholder	X2
R&D Investment	Ratio of R&D investment to main business income × 100%	X3
Percentage of Employees with Bachelor’s Degree or Above	Ratio of employees with bachelor’s degree or above to total number of employees × 100%	X4
Government Grants	Government subsidies (yuan)	X5
Management Skills	Management expense/main business income × 100%	X6
Return on Equity (ROE)	Net profit/total assets × 100%	X7
Total Asset Turnover Ratio	Net operating income/average total assets × 100%	X8

. Descriptive statistical analysis of the comprehensive technological innovation efficiency and the eight indicators mentioned above is shown in

Table 8. Descriptive statistical analysis of grey relational analysis indicators.

Variable	Quantity	Min	Max	Average	STDEV
Comprehensive Technological Innovation Efficiency	104	0.110	1.000	0.624	0.313
Total Assets	104	741,926,500.000	849,333,279,599.190	80,725,528,782.239	157,925,660,164.157
Shareholding Ratio of the Largest Shareholder	104	0.211	0.743	0.468	0.163
R&D Investment	104	0.010	0.126	0.046	0.024
Proportion of Employees with a Bachelor’s Degree or Above	104	0.072	0.814	0.327	0.175
Government Grants	104	187,780.140	14,027,129,065.720	1,206,020,239.639	2,738,449,574.105
Management Skills	104	0.014	0.208	0.074	0.042
Return on Equity (ROE)	104	−0.059	2.330	0.158	0.242
Total Asset Turnover	104	31.700	229.706	3.187	23.239
Significant Figure	104

.

Grey relational analysis was conducted on 104 data points across eight sub-factors (total assets, shareholding ratio of the largest shareholder, R&D investment, proportion of employees with bachelor’s degree or above, government subsidies, management ability, return on net assets, and total asset turnover rate). The technological innovation efficiency, calculated using the two-stage dynamic network DEA model, was used as the parent factor and its correlation with the eight sub-factors was studied. Grey relational analysis was performed with a resolution coefficient ρ = 0.5. The correlation coefficient was calculated using the relevant formula, and the correlation degree was then calculated based on the correlation coefficient value. The correlation degree results are shown in

Table 9. Results of grey relational analysis (GRA).

Indicator	Relevance	Sort
Total Assets (X1)	0.712	8
Shareholding Ratio of the Largest Shareholder (X2)	0.997	1
R&D Investment (X3)	0.994	5
Proportion of Employees with a Bachelor’s Degree or Above (X4)	0.996	2
Government Subsidies (X5)	0.712	7
Management Ability (X6)	0.994	4
Return on Equity (X7)	0.995	3
Total Assets Turnover Ratio (X8)	0.977	6

The correlation degree ranking is as follows: r₂ > r₄ > r₇ > r₆ > r₃ > r₈ > r₅ > r₁.

.

The correlation coefficients for the shareholding ratio of the largest shareholder, R&D investment, the proportion of employees with a bachelor’s degree or above, management ability, return on net assets, and total asset turnover rate are all above 0.990. These results indicate that these six factors are the most important factors affecting the comprehensive technological innovation efficiency of listed new energy vehicle companies.

The correlation coefficient between the shareholding ratio of the largest shareholder and the comprehensive technological innovation efficiency is the highest, reaching 0.997. As the shareholding ratio of the largest shareholder increases, their involvement in business operations becomes more proactive, thereby exerting a greater influence on innovation efficiency.

The second highest correlation degree is 0.996 for the proportion of employees with a bachelor’s degree or above. An increase in the proportion of employees with a bachelor’s degree or above has a positive effect on the overall technological innovation efficiency of enterprises. Enterprises with greater knowledge reserves have more talent with strong learning and innovation abilities. Employees with higher education levels have a relatively broader knowledge base, which enhances their ability to discuss and analyze problems during research and development. These factors facilitate communication and learning among scientific researchers, improve the atmosphere of innovation activities, and promote enterprise innovation levels.

The correlation coefficient of the return on net assets is 0.995, indicating that this factor reflects the efficiency of business operations and potential for future growth; it is also particularly useful for comparing profitability within the same industry.

The correlation coefficient of management ability is 0.994, indicating that efficient enterprise management can promote the development of employee efficiency and enhance enterprise benefits.

The correlation coefficient of R&D investment is 0.994, indicating that an increase in R&D investment can promote a company’s technological level and product quality, playing an important role in its innovation process.

The correlation coefficient of total asset turnover is 0.977, indicating that a higher total asset turnover promotes the generation of innovative activities and improves the overall efficiency of technological innovation.

The correlation coefficient of government subsidies is only 0.712. The technical threshold for fiscal subsidies for new energy vehicles in China is relatively low, and thus there are large amounts of subsidy funds with a low efficiency in fund allocation, which is not conducive to independent innovation. In the early stages of industrial development, there is a shortage of funds, and a country needs to invest a large amount of money in this area. Fiscal subsidies are conducive to technological research and development in enterprises. However, as the amount of subsidies increases, fiscal subsidies have a crowding-out effect on enterprises’ own R&D investment, which is not conducive to technological innovation. Thus, the role of government subsidies in enterprise technological innovation exhibits an “inverted U-shaped” pattern. On the other hand, due to the low threshold and single standard of fiscal subsidies, the subsidy standard for new energy passenger vehicles is determined by the endurance mileage. This subsidy standard is too simple, and the threshold is relatively low, leading to excessive dependence of new energy vehicle enterprises on fiscal subsidies, which is also not conducive to independent innovation. Faced with these problems in the context of the development of new energy vehicles, the government has decided to gradually reduce relevant policy incentives.

The correlation coefficient of total assets is only 0.712. As the size of a large enterprise increases, it becomes more difficult to manage, leading to lower efficiency. This can also reduce its sensitivity to market changes and hinder its ability to innovate.

5. Conclusions

5.1. Conclusions

Using a two-stage dynamic network DEA model, this study analyzed the comprehensive efficiency and stage efficiency of 13 listed new energy vehicle companies from 2017 to 2024, conducting GRA to analyze their influencing factors. The conclusions are as follows:

• The overall efficiency of the new energy vehicle industry is not high. The model results indicate that the industry’s overall efficiency reduced during the study period, highly related to the external environment, including policies, laws, government subsidies, environmental protection requirements, the emergence of new technologies, and policy instability and unsustainability. By the end of 2021, China had become the world’s largest producer and seller of new energy vehicles, and related technologies have continued to make progress. However, in the process of implementing relevant industrial policies, many problems have emerged, such as frequent fraudulent subsidies and the excessive dependence of enterprises on the dividends brought by industrial policies. This has led to insufficient R&D investment by the enterprises themselves, resulting in low technological innovation efficiency. Faced with these problems and the development situation of new energy vehicles, the government has decided to gradually reduce relevant policy incentives. The drastic fluctuations during the study period may be due to changes in external factors driving changes in internal factors within enterprises, for example, internal reforms, equity changes, institutional changes, and organizational changes.
• The comprehensive and stage efficiencies of listed new energy vehicle companies are not high. Among the 13 sample companies, only 2—Aote Xun and BYD—have a comprehensive efficiency value of 1; 3—Aote Xun, BYD, and Dongfeng Motor—have a technology research and development stage efficiency value of 1; and 4—Aote Xun, BYD, Dongfeng Motor, and Zhongtong Bus—have a results transformation stage efficiency value of 1. These companies thus exhibit technical effectiveness. At the same time, the efficiency of the technology research and development stage is higher than that of the commercial transformation stage, while the efficiency of technological innovation varies greatly depending on the nature of the enterprise. The technological innovation and stage efficiencies of state-owned enterprises are significantly higher than those of private enterprises. The main reason for this is that state-owned enterprises undertake more basic research as well as key development and technological innovation research. State-owned enterprises have become pioneers in the fields of new energy, medicine, electronic information, semiconductors, aerospace, high-speed rail, etc. These enterprises can also obtain more support from national-level policies and technologies.
• The six most important factors affecting the technological innovation efficiency of listed new energy vehicle companies are the shareholding ratio of the largest shareholder, R&D investment, the proportion of employees with a bachelor’s degree or above, management ability, return on net assets, and total asset turnover rate. On the other hand, government subsidies and total assets have a smaller impact on technological innovation efficiency.

5.2. Suggestions

• Government Level

The government should fully leverage its role in promoting and guiding innovation, implementing measures such as financial and innovation subsidies and tax and research and development incentives to enhance corporate enthusiasm for innovation. When formulating subsidy measures for new energy vehicles, the government should strengthen its supervision and management of listed companies and intensify penalties to reduce operational violations and prevent subsidy fraud. During the promotion of new energy vehicles, challenges such as imbalances in research and development capabilities among enterprises and significant disparities in benefits persist. The government should facilitate technology transfer between enterprises to help improve the production efficiency of small- and medium-sized enterprises. The government should encourage enterprises to expand their scale through mergers and reorganizations, avoiding homogenization and leveraging innovative resources to enhance technological innovation efficiency. Additionally, the government should allocate infrastructure investments across the eastern, central, and western regions. The national government should collaborate with enterprises, power companies, local governments, and other relevant departments to establish a network layout for new energy vehicle charging equipment, promoting collaborative innovation to drive the development of the entire industry.

2.

Industry Level

In the past decade, Chinese new energy vehicles (NEVs) have been in a stage of rapid development, and can overtake on corners and change lanes on lithium-ion batteries. In the NEV industry, low-carbon, energy-saving, and environmental protection approaches should be highlighted in multiple areas—such as raw materials, research and development, manufacturing, sales, and application—to form a closed supply–value–industry–industrial chain loop. NEV enterprises can implement diversified development strategies. While developing pure electric, renewable energy, and hybrid electric vehicles, enterprises should also strengthen the integration of new technologies, such as the internet, informatization, big data, and intelligence, and offer a variety of products to expand their development. At the same time, they should pay attention to brand-building and brand awareness for NEVs, develop competitive NEV products with well-known brands, and form dominant and chain-leading enterprises. They should also strengthen the intellectual property protection of NEVs, learn from international legal protection policies for innovation achievements according to the local NEV development needs, formulate and improve relevant domestic laws and regulations for technological innovation, provide more guarantees for the development of NEV enterprises, improve their innovation efficiency, build a collaborative innovation mechanism for NEVs, improve industrial development, and establish a cross-regional, cross-enterprise, and cross-departmental collaborative innovation mechanism. State-owned enterprises are the main body of basic technological innovation. While leveraging the technological innovation advantages of state-owned enterprises, we should also actively mobilize private enterprises’ enthusiasm for technological innovation, optimize and integrate various resources (such as products, technology, and human resources) within enterprises, and take full advantage of each enterprises’ strengths to effectively integrate the technology, industry, capital, and value chains.

3.

Corporate Level

New energy vehicle enterprises should increase their investment in research and development and intensify their efforts in developing new products and technologies. In the development of these enterprises, highly educated talent should be fully utilized. A talent cultivation mechanism should be established, and these enterprises should cooperate with universities to establish school–enterprise cooperation, cultivate and provide talent for their development, and improve their professional level. They should also better protect intellectual property rights, establish innovative research and development platforms, promote the integration of research and development findings into the market, form high-end international technical teams, break through key technological bottlenecks, attract key talent, and enhance the overall innovation vitality.

To enhance their technological level and management ability, enterprises should track the overall development of the new energy vehicle industry, strengthen technological innovation work, and formulate technology development strategies. The new energy vehicle industry should make structural adjustments; upgrade and optimize the industry; follow innovation pathways; strengthen technology research and development, technology promotion, and other services; produce high-quality, low-cost products; meet the needs of market economic development; and transform and upgrade technologies. At present, the core technology of new energy vehicles has defects; the problem of short-distance driving still persists. To promote the sustainable development of new energy vehicles, enterprises should strengthen their independent research and development efforts, break through core technologies, reduce external dependence, establish technological innovation laboratories, strengthen technological breakthroughs, and look for breakthroughs in core technologies such as batteries, electronic control, and motors. To enhance their management ability, enterprises should create an effective talent selection system to identify high-quality management candidates, improve their internal training mechanisms and reward/punishment mechanisms, and encourage management of managers through an appropriate salary system.