Factors Influencing the Acceptance of Electric Ride-Hailing Vehicles and Market-Exit Decisions Among Drivers in Underdeveloped Cities

Abstract

This study investigates the factors influencing electric ride-hailing vehicle (ERV) adoption and market exit among ride-hailing drivers in underdeveloped cities, aiming to enhance industry operational efficiency and workforce stability. Using survey data from 630 drivers in Zhangzhou, Fujian Province, China, we employ statistical modeling to analyze heterogeneous mechanisms behind ERV acceptance and market-exit behavior. The results indicate that: (1) Drivers’ environmental awareness and positive attitudes toward new energy vehicle development significantly increase ERV adoption willingness, though weak policy implementation and misconceptions about electric vehicle technology are associated with these incentives. Unlike in developed regions, range anxiety is not a significant barrier in underdeveloped cities with limited operational ranges. (2) Concerning market-exit behavior, job intensity exhibits a U-shaped relationship with exit risk, while platform income redistribution adjustments emerge as a key external factor. Notably, local full-time drivers show a marked lag in technology adoption. The findings offer valuable insights for optimizing policies and promoting sustainable development in the ride-hailing industry in underdeveloped cities, providing practical guidance for policymakers and companies in shaping tailored industry regulations and operational strategies.

1. Introduction

Accelerating global climate governance, the transportation sector, the major contributor to carbon emissions, stands as a pivotal avenue towards achieving the climate goal outlined in the Paris Agreement. As one of the world’s largest emitters, China has pledged to peak CO₂ emissions by 2026 and to achieve carbon neutrality by 2060, as the “dual carbon” strategic goals (carbon peaking and carbon neutrality) were planned. The International Energy Agency reports that the transport sector accounts for approximately 27% of global net greenhouse gas emissions, with road transport responsible for over 75%. This underscores the necessity of decarbonizing the transportation sector, particularly road transport, to meet the strategic goal of dual carbon [1,2]. Within this framework, electric ride-hailing vehicles (ERVs), as a novel mobility paradigm, utilize intelligent platform-based scheduling to improve operational efficiency and create numerous employment opportunities. Contrasted with the traditional fuel-powered ride-hailing vehicle with the average annual carbon emissions of 4.6 tons, pure electric vehicles (PEVs) reduce emissions by 85.71 kg monthly during operation [3].

Facing energy and environmental pressures, ERVs are gaining global adoption. Their appeal lies in cleanliness, flexibility, and convenience. The essence of the effectiveness of promoting electromobility is the collaborative product of technical equipment and social acceptance, which involves multiple dimensions such as vehicle usage, technical and operational plans, infrastructure, economic and legal issues [4]. Developed cities represented by Shenzhen have accomplished 100% electrification of ride-hailing vehicles through compulsory legislative measures, exemplifying urban development. Nevertheless, in underdeveloped areas, e-mobility infrastructure is an important factor restricting the adoption of ERVs. Data also indicate that the penetration rate of ERVs in underdeveloped urban areas stands at a mere 25%, significantly lower than the 46% average observed in developed cities. This pronounced regional disparity highlights a structural barrier in the ongoing electrification transition, jointly constituted by inadequate technological infrastructure and drivers’ acceptance. Drivers in underdeveloped areas exhibit distinct traits, including higher Hukou (local household registration) and lower educational attainment, that shape their decision-making. Thus, their choices are further influenced by regional economic constraints, infrastructure gaps, and limited social support systems, creating unique adoption barriers absent in developed cities. Nevertheless, existing research predominantly focuses on drivers’ acceptance of ERVs in developed cities. This oversight limits the understanding of the overall developing trends in the ride-hailing industry and hinders the creation of customized marketing plans for ERVs in underdeveloped regions.

Building on this foundation, this study focuses on drivers in underdeveloped regions. It investigates the acceptance of ERVs among both active and potential drivers (individuals licensed for ride-hailing but not currently driving), and examines the mechanisms influencing their exit from the ride-hailing market (no longer engaged in the ride-hailing driver profession). The innovation points of this study mainly lie in the following three aspects: first, using the ordered logit model to conduct an in-depth analysis of electric ride-hailing drivers and potential users, accurately identifying the factors influencing their acceptance, and systematically comparing the similarities and differences between the two groups; second, systematically analyzing the influencing factors of ride-hailing drivers exiting the ride-hailing market, providing new insights into drivers’ career-choice behavior; third, based on the research findings, proposing targeted policy recommendations to enhance drivers’ acceptance of ERVs, stabilize the ride-hailing market in underdeveloped cities, and offer strong theoretical support for the government, vehicle manufacturers, and ride-hailing vehicle rental platforms to formulate scientific and reasonable policies.

The remainder of this paper is structured as follows. Section 2 reviews the pertinent literature on electric ride-hailing drivers and their adoption determinants, highlighting the existing gaps within underdeveloped urban contexts. Section 3 delineates the empirical strategy, including the data collection approach, study area, variable construction, and respondent characteristics. Section 4 presents the research methodology and model specifications. Section 5 discusses the key findings of the empirical analyses on ERV acceptance and market-exit determinants. Section 6 derives targeted policies and recommendations from multiple perspectives based on the research outcomes. Section 7 briefly concludes the paper.

2. Literature Review

While ride-hailing services have been in operation globally for an extended period, the focus on electrification has emerged more prominently in recent years. According to the search results of the existing literature, the research on the acceptance of electric ride-hailing drivers is still limited. In this paper, we have systematically sorted out the differences in characteristics between drivers of electric and traditional ride-hailing vehicles, as well as the impediments and facilitators affecting the acceptance of ERV, aiming to provide a theoretical basis for subsequent related research.

2.1. Characteristics of Electric and Traditional Ride-Hailing Drivers

The characteristics of ride-hailing drivers help to reveal the key factors affecting the promotion of ERV and provide a theoretical basis for the development of differentiated policies. Xu et al. [5] empirically studied drivers’ response behavior to ride-hailing requests based on data from ride-hailing platforms, and clarified that drivers are more likely to respond to requests with economic incentives. Utilizing ride-hailing and taxi order data from Xiamen, Xiong et al. [6] categorized driver traits under different operational models: full-time ride-hailing drivers exhibit lower work intensity and stability compared to traditional taxi drivers, while part-time ride-hailing drivers work irregular hours, often concentrating on peak morning periods to supplement income. Sanguinetti and Kurani [7] revealed the characteristics and experiences of electric ride-hailing drivers through a survey of the Uber ride-hailing platform, and found that users of pure electric vehicles and plug-in hybrid electric vehicles (PHEVs) had concerns about range, which affected their experience and vehicle selection, with PEV users prioritizing expanded charging infrastructure. Chen et al. [8] investigated factors contributing to smoking behaviors among Chinese ride-hailing drivers, revealing a significant proportion of smokers—a reflection of the industry’s work stress. Chen et al. [9] investigated the effects of distractions in different ride-hailing systems on the driving performance of taxi drivers, which affect driving activity and safety. Kang et al. [10] addressed equity concerns in ride-hailing services, including earning differences among drivers and variance in passenger waiting times, highlighting their economic and ethical implications.

2.2. Factors Influencing Electric Ride-Hailing Adoption

The significance of studying the influencing factors for the acceptance of ERVs lies in clarifying the key obstacles restricting the electric transformation of the ride-hailing industry and providing a scientific basis for the accurate formulation of regionally differentiated promotion policies. Existing research classifies the acceptance of electric vehicles (EVs) more in the context of private car purchases. For instance, Li et al. [11] categorized the influencing factors into demographic, situational, and psychological factors. Furthermore, although these classifications cover individual and fundamental scenario factors, they pay insufficient attention to dimensions such as the unique social interactivity and operational attributes of ERV [12]. Based on this, this paper combines the social service attributes and sustainable transformation goals of ERVs and proposes a new classification framework: personal attributes, ethical norms, social characteristics, and vehicle characteristics. This classification not only integrates the core factors of existing research but also more accurately fits the scenarios of the ERV through new dimensions. The vehicle characteristics highlight the weight of operating attributes such as driving range and operating cost in decision-making [13]. Integrating the platform, policies, and society with social characteristics can reflect the social externalities of shared mobility [14,15]. This classification provides a more focused analytical framework for subsequently identifying the key factors of market-exiting decisions.

2.2.1. Personal Attributes

Many studies have classified gender, age, occupation, and education level as individual factors [16,17]. They pointed out that the group that shows a strong inclination towards EV adoption usually has a relatively high level of education and economic capacity. After examining Budapest residents’ post-implementation of urban vehicle access regulations, Ogunkunbi and Meszaros [18] indicated age (35–44 years) to be the strongest demographic predictor of battery electric vehicle (BEV) adoption, with household attitudes showing negligible influence. Du et al. [19] investigated how drivers’ registered permanent residence in China affects electric ride-hailing adoption, offering a novel lens on acceptance barriers and facilitators. Langbroek et al. [20] revealed that individuals leasing EVs exhibited higher acceptance and deeper understanding of EV technology.

2.2.2. Ethical Norms

Environmental issues directly affect people’s attitudes towards EVs [21,22]. Consumers will pay extra for EVs to protect the environment [23]. Its essential function is to activate the internalization process of an individual’s moral norms. Adnan divided the variables of personal ethics into self-transcendence, self-interest, openness to change, conservation, principles, and self-discipline. Environmental damage, technological innovation, innovation enjoyment, and technical knowledge. He and Zhan [24] demonstrated that personal norms have an optimistic effect on the adoption of EVs, which is influenced by external costs. When personal norms related to the environment are activated, personal moral responsibility is formed, which then guides individuals to perform pro-environmental behaviors. Focusing on Assam, India, Deka et al. [25] identified social circle adoption, self-efficacy in EV operation, and intrinsic moral responsibility toward sustainability as key adoption drivers. Viola [26] analyzed EV acceptance across regions, noting that adopters typically possess higher education levels, stronger environmental consciousness, greater financial capacity, and openness to new technologies.

2.2.3. Social Characteristics

The impact of social characteristics on the penetration of EVs in the online car-hailing industry is mainly reflected in two aspects: social incentives and government support [17]. Social incentives (fuel prices, platform commission rules) are closely related to drivers’ operating costs, while government support (infrastructure, subsidies, taxes, etc.,) provides guarantees for choosing electric ride-hailing vehicles. Adopting a fleet management perspective, Sugihara and Hardman [27] revealed that governmental and societal incentives—such as purchase subsidies, tax breaks, low-emission zone access, and restrictions on internal combustion engine vehicles—significantly shape EV adoption. Tu et al. [28] quantified the feasibility, energy consumption, and costs of electric ride-hailing fleets using GPS trajectory data from Beijing drivers. Their findings highlighted divergent acceptance levels among drivers under different service models and underscored insufficient charging infrastructure as a critical adoption barrier. Silva and Lampo [29] demonstrated that subjective perceptions and marketing strategies decisively influence purchasing decisions. Hackbarth and Madlener [30] analyzed consumers’ preferences for alternative fuel vehicles and found that consumers were more willing to accept vehicles with low fuel costs and fewer emissions, which indirectly reflected the relationship between acceptance and fuel prices. Gail Helen Broadbent [31] obtained the key role of the Australian government’s policy support, such as infrastructure deployment and vehicle rebates, in the popularization of EVs in cities by establishing a dynamic model.

2.2.4. Vehicle Characteristics

Vehicle attributes refer to the inherent physical and technical features of EVs and the manifestations of their directly derived usage effects, which together constitute the core characteristics that people face when purchasing, operating, maintaining, and experiencing such vehicles [32]. Saputra et al. [33] analyzed the influence of factors such as attitude and risk perception on the intention to adopt PEVs, and showed that safety risks have a significant negative impact on the willingness to adopt PEVs. Moreover, multiple studies corroborate that range anxiety [34], charging duration [35], and charging costs [36] remain primary consumer concerns hindering EV uptake, which seriously hinders the popularity of EVs. In addition, there is another perceived risk related to health, namely the risk of electromagnetic radiation. Although studies have shown that the EMF exposure level of EVs is comparable to or lower than that of many household appliances, consumers may still be concerned and consider it a risk [37,38].

Current research has established a comprehensive framework for studying the factors influencing consumer acceptance, purchasing behavior, and usage patterns of electric vehicles. However, this research is geographically limited, predominantly concentrating on developed regions and specific first-tier cities. Notably, there is a lack of systematic research regarding the decision-making process for the adopting and the withdrawal behaviors of electric car-hailing drivers in underdeveloped cities. These regions exhibit distinct characteristics such as development stage, infrastructure, income levels, and consumption patterns, which differ from those in developed areas. Consequently, existing research findings may not be directly applicable in these contexts. Moreover, existing studies have primarily concentrated on either driver characteristics or factors influencing the acceptance of ERVs, without integrating the two aspects. Examining the willingness to adopt and the behavior of exiting from the ride-hailing market among drivers in underdeveloped areas is crucial. This approach not only sheds light on the unique dynamics of electric ride-hailing adoption in such regions but also offers a fresh theoretical perspective and practical insights for promoting sustainable transportation transitions in developing nations.

3. Data

3.1. Study Area

Zhangzhou City, located in the south of Fujian Province, China, is a typical third-tier city that occupies a unique position in the regional economic pattern. According to the “Statistical communiqué of National Economic and Social Development in Zhangzhou in 2023”, by the end of 2023, the city’s permanent resident population had reached 5.063 million, with 3.25 million urban permanent residents. The private car park in the city is 709,300, and the sedan park is 446,000. There are 23 bus lines in Zhangzhou, which transport about 106,000 passengers per day [39]. In the development process of the ride-hailing industry, more than 4000 new people successfully obtained an “Online ride-hailing Driver’s License” in 2023 [40].

In 2023, the “Implementation Rules for the Management of Online ride-hailing Operation and Service in Zhangzhou City” clearly stipulates that passenger cars with fewer than 7 seats need to use new energy vehicles in principle [41]. By 2024, more than 1200 new and updated taxi (online ride-hailing) vehicles had been added, and the proportion of new energy vehicles reached 100% [42]. However, behind the booming development of the ride-hailing industry, there are also hidden worries. In 2023, about 8924 people applied for the qualification exam for online car practitioners, but the absence rate was as high as 24.92%, implying the proportion of drivers exiting the market was higher [40].

In this study, Zhangzhou City of Fujian Province is selected as the empirical research object, and its typicality is reflected in three aspects: first, as a regional central city with a permanent population exceeding 5 million and an urbanization rate of 64.2%, its urban scale and transportation structure highly represent third-tier cities in China. Second, the penetration rate of new energy vehicles and the ride-hailing market are typically representative. In 2023, the new energy rate of ride-hailing vehicles reached 100%, but the absence rate of the qualification examination of employees in the same period reached 24.9%, revealing that there is a significant gap between the efficiency of policy implementation and market response. Third, as an important node city in the Western Taiwan Straits Economic Zone, its blend of mid-size scale, moderate development, and strong low-carbon policies makes it an ideal microcosm to study sustainable transport challenges in developing regions, where these factors jointly shape unique transition dynamics. Therefore, a comprehensive analysis of the transportation characteristics in Zhangzhou City holds significant implications for understanding the electric ride-hailing market in underdeveloped urban areas.

3.2. Variable Identification

Most of the existing studies focus on the developed cities or the general ride-hailing market, and there are still insufficient studies on the acceptance of ERVs by drivers in underdeveloped cities and their exit from the market. Underdeveloped cities exhibit unique traits—such as lower income levels and relatively imperfect infrastructure—that may significantly influence drivers’ attitudes and decisions toward ERV. So this study employs a questionnaire survey to deeply investigate ERV acceptance and exit intentions among drivers in underdeveloped cities. In the questionnaire design, careful consideration was given to both the research objectives and the characteristics of the variables. For the dependent variables, acceptance of ERVs is represented by five-level indicators (strongly accept = 5, accept = 4, neutral = 3, reject = 2, strongly reject = 1). Market-exit behavior is represented by two-level indicators (entry = 1, exit = 2) for intuitive reflection of drivers’ decisions. Independent variables cover personal attributes (such as age, income, driving experience, etc.), vehicle and driving characteristics (such as vehicle type, daily driving hours, etc.), vehicle characteristics (such as range, charging duration, etc.), infrastructure (such as distribution of repair facilities, charging convenience, etc.), and attitudinal perception information (cognition of ERVs, attitude toward environmental protection, etc.). Details of the variables are presented in

Table 1. Variable categories.

Factor Category	Factors
Personal attributes	Occupation, gender, age, driving experience, Hukou (registered permanent residence), current vehicle type (including hybrid), prior use of gasoline vehicles, education level, work intensity, income
Personal ethical norms	Environmental awareness, attitude toward shared e-bike adoption, attitude toward EVs’ development impact on urban public transport
Social characteristics	Policy, ride-hailing infrastructure in region, platform commission rates, fuel price fluctuations, difficulty obtaining dual ride-hailing certificates (online ride-hailing vehicle operating license and qualification certificate)
Vehicle characteristics	Range, charging efficiency/time, smart features (autonomous driving), safety, comfort/experience, maintenance costs, insurance costs, driving operability, potential health impact of EV radiation

.

3.3. Data Source

The study was conducted using online and offline surveys from 13 November to 1 December 2024. For the offline component, the research team actively collaborated with relevant institutions in Fujian Province. After elucidating the purpose of the questionnaire survey to responsible personnel and securing their recognition and support, questionnaires were distributed at major government service windows and ride-hailing driver examination halls throughout Fujian Province. A questionnaire was distributed at the government service window to enhance the trust of respondents to the questionnaire with the help of its public credibility; in the online ride-hailing examination hall, drivers who are ready to engage in or have engaged in the online ride-hailing industry can be directly confronted to ensure the professionalism and pertinence of the survey samples. Simultaneously, team members utilize social media to precisely target and extensively recruit driver users for survey participation.

To accurately identify the target population, a series of focused strategies has been implemented. First, questionnaires were distributed through high-activity online communities frequented by ride-hailing drivers, including WeChat’s discussion groups and Weibo’s super topics. Platform algorithms further enhanced targeting by recommending the survey to users demonstrating interest in ride-hailing content based on their browsing history and preferences. The second is to carefully set screening questions in the initial part of the questionnaire. Only users who give qualified answers to these questions can enter the main part of the questionnaire to fill in, thereby effectively filtering out non-target groups and improving the efficiency and quality of data collection. Subsequently, users completing the survey undergo preliminary screening. Through analysis of users’ multi-dimensional information like browsing history, areas of interest, and interactive patterns, discrepancies from the behavioral traits of online taxi drivers are accurately identified to be verified and addressed during subsequent data cleaning processes.

This study employs rigorous quality control measures to ensure survey data reliability. Prior to the formal survey, investigators received systematic training on questionnaire design principles, technical terminology interpretation, and survey implementation specifications. Following a pre-test involving 15 professional drivers, the questionnaire underwent optimization and adjustment to establish a 4 min completion time threshold, serving as a basis for subsequent data quality control. During the formal survey, a standardized process was implemented: investigators first explained the purpose of the survey to participants before guiding them through questionnaire completion. To enhance data quality, drivers received CNY 3 as a reward for valid submissions, and an incentive mechanism encouraged questionnaire sharing to expand sample coverage. For data processing, a consistent online and offline questionnaire design approach was adopted to minimize measurement bias. Data cleaning involved applying dual screening criteria for response time (<4 min) and logical consistency, with manual review of anomalous data. Additionally, random sampling enabled cross-validation between online and offline responses, ensuring data consistency and reliability.

Ultimately, 724 questionnaires were collected. A rigorous data-cleaning process removed samples with logical errors or excessively short response times. After meticulous screening, 630 valid samples were retained, representing 87.02% of the total collection, laying a solid data foundation for the subsequent analysis of drivers’ acceptance of electric network vehicles and the driving factors for exiting the market.

3.4. Basic Attribute Analysis of Respondent Characteristics

In the analysis of respondent attributes, full-time ride-hailing drivers account for 46.5%, indicating that the occupation has significant local appeal. In terms of gender distribution, male drivers account for a dominating 71.1%, while female drivers make up 28.9%, consistent with the “2024 Female Driver Employment Data Report” released by T3 Mobility [43]. This may be because the ride-hailing driver embraces more flexible working hours, meeting their desire to balance income and family.

In terms of educational background, the education level of drivers is diversified, with college education accounting for 31.9%, high school education accounting for 29.4%, reflecting that more than half have received higher education. The age distribution is highest in the 26–35 age group accounting for 46.8%. Additionally, 34.1% of the drivers have more than 10 years’ driving experience, which provides occupational stability and guarantees the promotion of electric ride-hailing. It is worth noting that 87% of the drivers are non-local residents, revealing the unique appeal of the industry to migrants.

There are 62.4% of active EVs in use, 53.7% of drivers work an 8–12 h day, and 39.5% of drivers earn CNY 5000–7500 (Chinese Yuan) monthly, reflecting the labor intensity and income stability of the industry. In addition, car purchase subsidies (40.3%) and operation subsidies (35.7%) became the most concerned support measures for drivers. These findings provide an important reference for the subsequent study of drivers’ acceptance of electric ride-hailing and the driving factors for their exit from the market (

Table 2. Statistical analysis of respondent attributes.

Variable	Attribute	Frequency	Percentage (%)
Occupation	Full-time ride-hailing driver	293	46.5
	Part-time ride-hailing driver	121	19.2
	Taxi driver	18	2.9
	Other	198	31.4
Gender	Male	448	71.1
	Female	182	28.9
Age	18–25 years	30	4.8
	26–35 years	295	46.8
	36–45 years	158	25.1
	≥46 years	147	23.3
Driving experience	≤3 years	72	11.4
	4–6 years	186	29.5
	7–9 years	157	24.9
	≥10 years	215	34.1
Hukou	Local	82	13.0
	Non-local	548	87.0
Current vehicle type	Driving non-EV	237	37.6
	Driving EV	393	62.4
Previous vehicle use	No prior use of gasoline vehicles	57	9.0
	Prior use of gasoline vehicles	573	91.0
Education level	Bachelor’s degree	163	25.9
	Middle school or below	65	10.3
	Associate degree	201	31.9
	High school diploma	185	29.4
	Master’s degree or higher	16	2.5
Work intensity	≤4 h/day	58	9.2
	4–8 h/day	202	32.1
	8–12 h/day	338	53.7
	≥12 h/day	32	5.1
Monthly income	≤2500 CNY	44	7.0
	2500–5000 CNY	226	35.9
	5000–7500 CNY	249	39.5
	7500–10,000 CNY	94	14.9
	≥10,000 CNY	17	2.7
Policy expectations	Charging subsidy	109	17.3
	Vehicle purchase subsidy	254	40.3
	Recycling subsidy	42	6.7
	Operational subsidy	225	35.7

).

4. Research Methodology

4.1. Conceptual Framework

Technology Acceptance Model

The Technology Acceptance Model (TAM), proposed by Davis et al. [44], is a theoretical framework developed from the Theory of Reasoned Action to explain individuals’ behavioral responses and attitudes toward adopting new technologies. The model posits that individual behavior is determined by behavioral intention, which in turn is influenced by one’s attitude toward new technologies or products and their perceived usefulness. This study strictly designs its analytical framework around the core variables of the TAM. Specifically, in the dimension of perceived usefulness (PU), the focus is placed on ride-hailing drivers’ perceived value of ERVs in terms of operational benefits (e.g., maintenance costs and insurance expenses) and the adaptability of ERVs’ driving range to local road conditions. These factors serve as key criteria for drivers to judge whether ERVs are “useful”. In the dimension of perceived ease of use (PEOU), the study primarily measures drivers’ perceptions of ERVs’ usage experience, which specifically includes “driving operation experience”, “whether the coverage of local charging piles meets operational needs”, and “experience with ERVs’ intelligent functions”. These indicators directly align with the core concept of “ease of use” in TAM.

Furthermore, to address this study’s focus on ERV adoption and market-exit behaviors among ride-hailing drivers in underdeveloped contexts, the paper incorporated several control variables into the framework: personal attributes such as age, gender, and income, which were tested for their influence on ERV adoption intentions. On the basis of comprehensive analysis, the integrated analytical framework also integrates a new policy dimension. As illustrated in the accompanying Figure 1, this integrated approach holistically elucidates the factors influencing ERV acceptance and market exit among ride-hailing drivers in underdeveloped areas, thereby enhancing the applicability and scientific rigor of the study.

4.2. Assessment

4.2.1. Reliability Assessment

Given that this study employed Likert scale-based statements in the questionnaire, a rigorous evaluation of the reliability was conducted prior to empirical analysis. Reliability was assessed using Cronbach’s alpha coefficient, with results detailed in

Table 3. Reliability statistics.

Dimension	Cronbach’s Alpha	Items
Vehicle characteristics	0.892	9
Total	0.887	29

. The obtained alpha is 0.887, indicating that the self-compiled questionnaire scale has a high reliability of questions, consistency, and stability, and is suitable for this survey.

4.2.2. Validity Assessment

Subsequently, a validity analysis was conducted to evaluate the accuracy and construct validity of the survey data. This study employed the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s test of sphericity for verification. The results, summarized in

Table 4. Validity analysis.

Measure		Value
Kaiser–Meyer–Olkin (KMO)		0.915
Bartlett’s test of sphericity	Approx. Chi-Square	8228.695
	Df	406
	Sig.	<0.001

Df = degree of freedom, Sig. = significance.

, demonstrate: the KMO statistic of 0.915 (exceeding the recommended threshold of 0.7), and Bartlett’s test significance at p < 0.001. These robust metrics confirm the questionnaire’s high validity and establish its appropriateness for subsequent factor analysis procedures.

4.2.3. Common Method Bias Assessment

Common method bias (CMB) refers to artificial covariation between predictor and outcome variables caused by identical data sources, raters, measurement contexts, or item characteristics. The present study, which relies on self-reported data, may be susceptible to CMB. To address the potential CMB in this research, we minimized and assessed the likelihood of their occurrence using procedural and statistical methods [45]. At the design stage, anonymous questionnaires were adopted to mitigate the social desirability effect, and data were collected via both online and offline channels to break the response inertia associated with a single data source, which helped reduce the potential risk of CMB. In terms of statistical validation, Harman’s single-factor analysis was first conducted for exploratory factor analysis (EFA). The results reveal eight factors with eigenvalues greater than 1, and the first factor accounted for 30.24% of the total variance—below the 40% threshold—indicating no significant CMB. To further verify the robustness of the model, confirmatory factor analysis (CFA) was then performed, and the model fit indices (χ² = 1230.182, df = 135, χ²/df = 9.112 > 3, RMSEA = 0.114 > 0.10, CFI = 0.805 < 0.9, TLI = 0.778 < 0.9) exhibited poor fit. Collectively, these findings indicate that CMB does not seriously threaten the validity of our results.

4.3. Model Construction

This study collected 630 valid samples through questionnaire surveys. Since the dependent variable (the acceptance of ERV) presents prominent characteristics of ordered categories, the ordered logit model was selected, enabling precise capture of ordered relationships between acceptance tiers and thorough exploration of hierarchical effects among influencing factors. The ordered logit model defines the dependent variable as an ordinal categorical variable representing acceptance levels.

$$y=\left\{\begin{array}{l}1\hspace{1em}(y^{\ast }\le {\mu }_1)\\ 2\hspace{1em}({\mu }_1\le y^{\ast }\le {\mu }_2)\\ 3\hspace{1em}({\mu }_2\le y^{\ast }\le {\mu }_3)\\ J\hspace{1em}({\mu }_{J-1}\le y^{\ast })\end{array}\right.$$

(1)

Since all probabilities are positive, therefore $0<{\mu }_2<{\mu }_3<\dots <{\mu }_{J-1}$, it is standard practice to fix the first threshold parameter ${\mu }_1$ at zero, resulting in $J-2$ parameters requiring estimation.

Expressed in logit form, the model can be expressed as the following:

$$\ln {\displaystyle \frac{P(y\le j\left|X\right.)}{1-P(y\le j\left|X\right.)}}={\mu }_j+{\displaystyle \sum _{k=1}^K{\beta }_k{x}_k}(j=1,2,\dots ,J-1)$$

(2)

Expressed as a probability:

$$\mathrm{Prob}(y\le j)=\mathrm{Prob}(y^{\ast }\le {\mu }_j)={\displaystyle \frac{e^{{\mu }_j-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k}}{1+e^{{\mu }_j-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k}}}$$

(3)

Replacing the equation generalized cumulative distribution function F with a cumulative distribution function representing a logistic distribution $L(\cdot )$, the ordered logit model is expressed in probabilistic form as follows:

$$\begin{array}{l}\mathrm{Prob}(y=1)=L\left(-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k\right)\\ \mathrm{Prob}(y=2)=L\left({\mu }_2-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k\right)-L\left(-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k\right)\\ \mathrm{Prob}(y=3)=L\left({\mu }_3-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k\right)-L\left({\mu }_2-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k\right)\\ \hfill ⋮\hfill \\ \mathrm{Prob}(y=J)=1-L\left({\mu }_{J-1}-{\displaystyle \sum _{k=1}^K{\beta }_k}{x}_k\right)\end{array}$$

(4)

For exit market choice, drivers have two options: either to exit (value 0) or remain (value 1) in the market. The binary logit model can be used to efficiently identify the key factors affecting drivers’ exit decision. The binary logit model expression is as follows:

$$\mathrm{Logit}(P)=\ln \left({\displaystyle \frac{P}{1-P}}\right)=\alpha +{\beta }_0+{\beta }_1{X}_1+\dots +{\beta }_k{X}_k+\epsilon$$

(5)

where $P$ denotes the probability of a driver exiting the market, $P/(1-P)$ represents the odds ratio of continued participation versus exit, $\alpha$ is the intercept, is the regression coefficient for independent variables $X_j$, and $\epsilon$ is the random error term.

4.4. Variable Calibration

Independent variable calibration was conducted before building the ordered logit model (analyzing ERV acceptance) and binary logit model (evaluating exit decisions). A total of 29 variables derived from the questionnaire—including occupation, gender, age, income, driving experience, environmental awareness, and attitudinal perceptions—are calibrated as shown in

Table 5. Factors influencing transportation mode choice and variable assignments.

Factor	Categories and Assignments
Occupation	Full-time ride-hailing driver = 1; part-time ride-hailing driver = 2; taxi driver = 3; other = 4
Gender	Male = 1; female = 2
Age	18–25 years = 1; 26–35 years = 2; 36–45 years = 3; ≥46 years = 4
Driving experience	0–3 years = 1; 4–6 years = 2; 7–9 years = 3; ≥10 years = 4
Local registered permanent residence	Yes = 1; no = 2
Current vehicle type	EV = 1; non-EV = 2
Prior use of gasoline vehicles	Yes = 1; no = 2
Education level	Middle school or below = 1; high school = 2; associate degree = 3; bachelor’s degree = 4; master’s degree or higher = 5
Environmental awareness	Strongly disagree = 1; disagree = 2; neutral = 3; agree = 4; strongly agree = 5
Work intensity	<4 h/day = 1; 4–8 h/day = 2; 8–12 h/day = 3; ≥12 h/day = 4
Monthly income (CNY)	≤2500 = 1; 2501–5000 = 2; 5001–7500 = 3; 7501–10,000 = 4; >10,000 = 5
Subsidy policy preferences	Operational subsidy = 1; recycling subsidy = 2; vehicle purchase subsidy = 3; charging subsidy = 4
Attitude Toward Shared E-bike Adoption	Strongly disagree = 1; disagree = 2; neutral = 3; agree = 4; strongly agree = 5
Attitude Toward EV Development	Strongly disagree = 1; disagree = 2; neutral = 3; agree = 4; strongly agree = 5
Evaluation of supporting infrastructure	Very poor = 1; poor = 2; neutral = 3; good = 4; excellent = 5
Satisfaction with functional attributes (Range, charging efficiency/time, smart features [autonomous driving], safety, comfort/experience, maintenance costs, insurance costs, driving operability)	Very dissatisfied = 1; dissatisfied = 2; neutral = 3; satisfied = 4; very satisfied = 5
Concerns about EV radiation impact on health	Very concerned = 1; concerned = 2; occasionally concerned = 3; rarely concerned = 4; not concerned = 5
Perceived fairness of platform commission rules	Very unfair = 1; unfair = 2; neutral = 3; fair = 4; very fair = 5
Impact of fuel price fluctuations	Severe impact = 1; moderate impact = 2; some impact = 3; low impact = 4; no impact = 5
Difficulty obtaining dual certificates	Very difficult = 1; difficult = 2; neutral = 3; easy = 4; very easy = 5
Acceptance of ERV	Very difficult = 1; difficult = 2; neutral = 3; acceptable = 4; very acceptable = 5
Decision behavior	Join = 0; exit = 1

.

4.5. Collinearity Test

Before conducting regression analysis, collinearity diagnostics were performed on all independent variables to mitigate potential bias in regression estimates resulting from high inter-variable correlations. This assessment employed the variance inflation factor (VIF) as the primary metric. As summarized in

Table 6. Collinearity test.

Variable	Tolerance	VIF
Occupation	0.4	2.502
Gender	0.862	1.16
Age	0.45	2.222
Driving experience	0.609	1.641
Local registered permanent residence	0.895	1.117
Current vehicle type	0.554	1.806
Prior use of gasoline vehicles	0.887	1.128
Education level	0.704	1.421
Environmental awareness	0.369	2.707
Work intensity	0.73	1.37
Monthly income (CNY)	0.739	1.353
Subsidy policy preferences	0.876	1.141
Attitude toward shared E-bike adoption	0.433	2.307
Attitude toward EV development	0.4	2.498
Evaluation of supporting infrastructure	0.469	2.132
Range	0.369	2.713
Charging efficiency/time	0.562	1.779
Smart features [autonomous driving]	0.539	1.856
Safety	0.303	3.305
Comfort/experience	0.371	2.693
Maintenance costs	0.41	2.437
Insurance costs	0.451	2.217
Driving operability	0.292	3.423
Concerns about EV radiation impact on health	0.648	1.543
Perceived fairness of platform commission rules	0.442	2.265
Impact of fuel price fluctuations	0.756	1.323
Difficulty obtaining dual certificates	0.671	1.491
Acceptance of ERVs	0.594	1.685
Mean VIF		1.973

, all VIF values in this study were substantially below the established threshold of 10 [46], indicating the absence of multicollinearity concerns. Consequently, the data satisfy the prerequisite conditions for rigorous empirical analysis.

5. Model Results and Discussion

5.1. Influencing Factors on ERV Acceptance

The ordered logit model results for the factors influencing the acceptance of electric ride-hailing vehicles among drivers in underdeveloped cities are presented below in

Table 7. Acceptance model results.

Variable	B	Sig.	Odds Ratio
Occupation = full-time ride-hailing driver	1.686	0.045 **	5.398
Occupation = part-time ride-hailing driver	−1.003	0.319	0.367
Occupation = taxi driver	−1.022	0.611	0.36
Gende = male	−0.951	0.006 ***	0.386
Work intensity ≤ 4 h/day	−0.92	0.076 *	0.399
Work intensity = 4–8 h/day	−0.711	0.09 *	0.491
Work intensity = 8–12 h/day	−0.462	0.258	0.63
Subsidy policy preferences = operational subsidy	0.45	0.078 *	1.568
Subsidy policy preferences = recycling subsidy	0.044	0.913	1.045
Subsidy policy preferences = vehicle purchase subsidy	−0.138	0.569	0.871
Decision behavior = exit	1.105	0 ***	3.018
Environmental awareness	0.573	0 ***	1.773
Shared E-bike adoption	0.38	0 ***	1.462
Attitude toward EV development	0.389	0 ***	1.476
Range	0.008	0.936	1.008
Charging efficiency/time	−0.11	0.24	0.896
EV radiation concerns	0.182	0.018 **	1.199
Perceived fairness of platform commission rules	−0.038	0.696	0.963
Impact of fuel price fluctuations	−0.062	0.461	0.939
Difficulty obtaining dual certificates	0.349	0 ***	1.418
Occupation = part-time * Gender = male	1.272	0.018 **	3.567
Occupation = full-time * Local residence = yes	−1.728	0.003 ***	0.178
Number of samples	630
Log-likelihood (final)	−609.031
Log-likelihood (null)	−932.561
AIC (final)	1342.062
AIC (null)	1873.121

* significance of 0.1, ** significance of 0.05, *** significance of 0.01. AIC = Akaike Information Criterion.

. All independent variables with a calibration value of 0 were set as reference items and were then selected with an inclusion probability of 0.05 and an exclusion probability of 0.1. With the Cox and Snell value of 0.642, the Nagelkerke value of 0.677, the McFadden value of 0.347 (greater than 0.2), the chi-square value of 647.059, the −2 log-likelihood value of 1023.894, and the significance of the parallel line test of 0.141 (greater than 0.05), the significance level of the final model was less than 0.05, proving that the results obtained from parameter estimation are accurate and reliable.

Referring to Table 7 and Figure 2 for detailed variable statistics, at the ethical norms level, environmental awareness (β = 0.571, p < 0.05), perceptions of shared e-bike popularization, and attitudes toward electric vehicle development significantly predict drivers’ acceptance of ERV. Each unit increase in environmental awareness corresponds to a 0.571 elevation in the log-odds of acceptance. Regarding electric vehicle perception, greater confidence in EV radiation safety associates with substantially higher acceptance likelihood, manifesting as a 1.199-fold odds increase per unit improvement (OR = 1.199). This relationship may stem from pervasive information asymmetries in underdeveloped areas, where misleading dissemination through media and social platforms substantially influences driver decision-making. Concerning personal attributes, full-time ride-hailing drivers demonstrate significantly stronger acceptance (OR = 5.398), indicating their odds of ascending one acceptance tier are 5.398 times greater than those of other driver categories. Conversely, male gender serves as an inhibitory factor (OR = 0.386), corresponding to a 61.4% reduction in acceptance escalation probability relative to females. Interaction analyses revealed counterintuitive patterns: local full-time drivers exhibited significantly depressed acceptance, with adjusted odds amounting to merely 17.8% of their non-local counterparts. Among part-time drivers, the male gender disadvantage was substantially attenuated, with males demonstrating 3.57 times higher acceptance odds than females, representing a contextual reversal of the primary gender effect observed in the broader sample.

5.2. Research on the Mechanism of Ride-Hailing Drivers’ Decision-Making to Exit the Market

5.2.1. Relationship Between Acceptance Level and Market-Exit Decision-Making Behavior

In the survey, a key investigation was conducted on drivers’ willingness to leave or not join the ride-hailing business. As shown in

Table 8. Relationship between all drivers’ acceptance levels and market-exit/-entry decision-making behaviors.

Acceptance Level	Exit	Entry	Total
Levels 1 and 2	35.58%	2.63%	19.68%
Level 3	23.01%	9.54%	16.51%
Levels 4 and 5	41.41%	87.83%	63.81%
Total	51.75%	48.25%	100.00%

, 326 drivers (51.75%) clearly expressed their intentions to exit the market. Among the drivers who are willing to continue in the ride-hailing business, as high as 87.83% had a high acceptance of electric ride-hailing vehicles, while only 2.63% had a low acceptance; in contrast, among those choosing to leave the business, 58.59% had a low acceptance. This indicates that the acceptance level of electric ride-hailing vehicles by drivers is significantly positively correlated with their willingness to remain in the business; that is, the higher the acceptance, the stronger the willingness to continue working; the lower the acceptance, the greater the possibility of leaving the business.

To more accurately analyze the relationship between willingness to exit the market and acceptance, as shown in

Table 9. Relationship between full-time ride-hailing drivers’ acceptance levels and market-exit/-entry decision-making behaviors.

Acceptance Level	Exit	Entry	Total
Level 1	36.65%	0.66%	19.17%
Level 2	22.98%	4.61%	14.06%
Level 3	23.60%	8.55%	16.29%
Level 4	6.83%	38.82%	22.36%
Level 5	9.94%	47.37%	28.12%
Total	51.44%	48.56%	100.00%

, the results reveal there is a significant linear relationship between the acceptance degree of professional ride-hailing drivers and the market-exit decision; this conclusion is particularly prominent among full-time drivers in underdeveloped regions. Full-time drivers with high acceptance have a significantly lower willingness to leave the business, while those with low acceptance have a significantly higher intention to quit.

5.2.2. Model Results of Market-Exit Decision-Making Behavior

Based on the relationship between acceptance and willingness to leave the business, the influence mechanism of ride-hailing drivers’ exit-decision behavior is analyzed, and the results are shown in

Table 10. Model results of drivers’ market-exit decision behavior.

Variable	B	Sig.	Odds Ratio
Occupation = full-time ride-hailing driver		0 ***
Occupation = part-time ride-hailing driver	0.138	0.759	1.147
Occupation = taxi driver	−1.416	0.107	0.243
Occupation = other	2.428	0 ***	11.333
Gender = female	−0.077	0.805	0.926
Current vehicle type = non-EV	0.991	0.003 ***	2.695
Prior use of gasoline vehicles = no	0.429	0.366	1.535
Environmental awareness		0.023 **
Work intensity		0.005 ***
Work intensity = ≥12 h/day	1.69	0.045 **	5.42
Subsidy policy preferences = operational subsidy		0.021 **
Subsidy policy preferences = charging subsidy	0.712	0.075 *	2.037
Attitude toward EV development = strongly disagree		0.138
Attitude toward EV development = disagree	−1.829	0.042 **	0.161
Attitude toward EV development = neutral	−1.241	0.138	0.289
Attitude toward EV development = agree	−1.856	0.012 **	0.156
Attitude toward EV development = strongly agree	−1.715	0.025 **	0.18
Concern about EV radiation impact on health = very concerned		0.001 ***
Concern about EV radiation impact on health = not concerned	−1.667	0.003 ***	0.189
Perceived fairness of platform commission rules = very unfair		0.021 **
Difficulty obtaining dual certificates = very difficult		0.232
Difficulty obtaining dual certificates = difficult	−3.138	0.037 **	0.043
Difficulty obtaining dual certificates = neutral	−2.692	0.07	0.068
Difficulty obtaining dual certificates = easy	−2.677	0.067	0.069
Difficulty obtaining dual certificates = very easy	−3.076	0.038 **	0.046
Log-likelihood (null)	−436.2985
Log-likelihood (full)	−228.279
AIC (null)	874.597
AIC (full)	640.558

* significance of 0.1, ** significance of 0.05, *** significance of 0.01. AIC = Akaike Information Criterion.

. With the Cox and Snell value of 0.483, the Nagelkerke value of 0.645 (far exceeding 0.3), the significance of the Hosmer–Lemeshow test of 0.329 (greater than 0.05), and the prediction accuracy of the regression model of as high as 84%, the final model significance level is less than 0.05, confirming that the selected variables reliably reflect true relationships within the dataset at the 95% confidence level.

Specific variables are shown in Table 10 and Figure 3. Regarding personal attributes, work intensity exceeding 12 h per day significantly increased the probability of market exit (OR = 5.42), indicating such drivers had 5.42 times higher exit likelihood than those working 0–4 h. Taxi drivers emerged as the predominant exit group (OR = 11.333), demonstrating 11.333 times greater exit odds than full-time ride-hailing drivers. Drivers without prior EV experience also exhibited high exit propensity, potentially attributable to entrenched reliance on fuel vehicles coupled with skepticism toward EV adoption. Within personal ethical norms, environmental awareness and attitude toward EVs’ development impact on urban public transport constituted key determinants of exit decisions. The low-environmental consciousness cohort (98.2% scoring at acceptance levels 1–2) showed heightened exit susceptibility, often due to rejecting EVs’ environmental value. Notably, positive attitudes regarding EVs’ development impact on urban public transport correlated negatively with market exit (p < 0.05): drivers holding favorable views demonstrated 5–6 times higher retention probability than the most negative cohort. For social characteristics, perceptions of difficulty obtaining dual certificates critically influenced retention: drivers considering licensing “difficult” or “very easy” showed significantly higher retention versus those reporting “very difficult” (p < 0.01). This suggests dual certificate accessibility optimization represents a pivotal policy for reducing driver attrition.

5.3. Discussion

After identifying the key factors influencing drivers’ acceptance of ERVs and their exit from the market in underdeveloped areas, this study will further analyze the driver behavioral patterns under specific circumstances. By comparing with existing mature conclusions in developed areas, this study aims to reveal both commonalities and differences in driver behaviors across different urban development stages.

In terms of acceptance levels, similar to developed regions [19,47,48], gender and occupational categories significantly influence drivers’ acceptance of ERVs in underdeveloped areas. However, distinct results emerge: full-time ride-hailing drivers show higher acceptance rates. This may stem from two factors: first, their longer exposure to EV information leads to greater technological acceptance; second, as primary household earners in economically underdeveloped areas with limited job opportunities, they are compelled to adapt to new policies. Contrary to previous research where males were more receptive to EVs, this finding in underdeveloped areas likely reflects delayed information dissemination. Their understanding of EV technology relies more on experiential judgment than professional data, while men’s heightened focus on vehicle performance and preference for traditional gasoline vehicles may further contribute to this disparity. Additionally, the immaturity of ERV markets in these regions intensifies risk-averse tendencies, a phenomenon particularly pronounced among part-time male ride-hailing drivers.

Policy incentives have demonstrated outcomes diverging from conventional expectations. While vehicle purchase subsidies and operational subsidies account for significant proportions in descriptive statistics, their statistical significance in ordered logit models remains modest. This finding contradicts Habich-Sobiegalla’s seminal assertion about policy incentives’ substantial impact on consumer purchasing decisions [49]. This discrepancy may stem from drivers’ reduced sensitivity to policies in underdeveloped regions, where vehicle pricing and brand factors may undermine subsidy effectiveness. Operational subsidies face implementation challenges due to regional market fairness constraints. Ethically, this study not only validates previous research on environmental motivations driving EV adoption [50,51] but also reveals the strong correlation between personal ethics and driver behavior in underdeveloped areas, expanding the geographical dimension of related studies. Within electric vehicle attributes, range has no statistically significant effect on ERV acceptance, contradicting findings from developed areas [7,34]. This may be attributed to limited operational scope and fewer orders in underdeveloped areas, where drivers’ daily work demands lower range requirements. In addition, groups more concerned with the potential health impact of EV radiation show greater resistance to ERVs. The widespread information asymmetry and misleading information from social media narratives prevalent in underdeveloped areas likely exert significant influence on drivers’ decisions.

Regarding market exit, studies reveal that drivers with a work intensity of 0–4 h or over 12 h exhibit an elevated probability of leaving the ride-hailing business. Drivers working 0–4 h have lower occupation stickiness, while drivers who work more than 12 h have health issues as a significant concern. Compared with developed cities, in underdeveloped regions with relatively limited medical resources, the imbalance between income and cost (time, treatment, transportation, and finance) has prompted some drivers to turn to more flexible and low-risk businesses [52]. Taxi drivers show higher exit rates, with data indicating about 40% of ride-hailing drivers receive fewer than 10 orders daily. Some abandon their jobs due to insufficient earnings, while government-mandated taxi pricing and strict license quotas further restrict operational autonomy, which is a key factor in their departure. Drivers unfamiliar with electric vehicles are more prone to make exit decisions, likely due to lingering distrust of EVs compared to traditional gasoline vehicles. Operational subsidy policies and platform commission rules significantly influence decisions. Recent declines in driver subsidies from 15–20% of revenue to 5–8% reflect capital expansion on platforms and inadequate driver protection [53], revealing involuntary reasons behind drivers’ exits.

Previous studies on driver exit from the market have predominantly focused on factors such as work flexibility, personal attributes, and platform characteristics [19,54]. This research establishes a connection between acceptance and exit decisions, revealing how individual ethical norms and social factors influence decision-making under new policy. Environmental awareness and attitudes towards the impact of EV development on urban public transportation systems are key factors influencing drivers’ decisions to exit. The difficulty in obtaining dual certificates for ride-hailing services shows a negative correlation with drivers’ decision-making behavior. Those perceiving licensing challenges as high tend to exit the market, a decision which may be linked to educational levels in underdeveloped regions and excessive risk expectations among drivers. This not only highlights issues in industry compliance but also underscores the importance of balancing operational safety standards with public needs in policy formulation. Governments should reconsider policies to create a competitive operating environment for electric vehicles [47].

6. Policy Implications and Suggestions

Based on the comprehensive analysis of the acceptance of ERVs and the underlying factors influencing exit decisions among drivers in underdeveloped cities, the following policies and suggestions are proposed.

1.

Optimize the Certificate Management System for Ride-Hailing Vehicles

The ease of obtaining dual certificates for ride-hailing services significantly influences drivers‘ adoption of electric vehicles. As shown in Table 7, each 10% reduction in license acquisition difficulty increases the probability of ERV acceptance by 41.8%, while reducing drivers’ likelihood of exiting the market by 95%. Therefore, a tiered licensing system should be established based on regional characteristics (with restricted operational areas for different tiers). For city-level vehicles, existing standards can remain unchanged; county-level vehicles may undergo simplified exam question banks with enhanced assessments of rural road safety knowledge. Additionally, taxi drivers require an extra “Taxi Driver License” compared to ride-hailing drivers, which creates operational challenges. This explains why taxi drivers exit the market 11.33 times more frequently than full-time drivers. To address this, the government could merge the exam question banks for both taxi and ride-hailing drivers, issuing licenses only to those passing the exams and meeting background checks [55].

2.

Regulate Platform Operations and Safeguard Driver Rights

Ride-hailing platforms and the government maintain a competitive relationship, constantly adjusting strategies through confrontation, interdependence, and regulation [56]. Full-time drivers must work 10 h daily to achieve an average income of CNY 300, creating an imbalance between “high input and low returns”. The platform’s order allocation rules and commission ratios further shape drivers’ decision-making. Therefore, platforms should be required to disclose real-time commission rates per order, enforce transparent regulations, and prohibit arbitrary adjustments to subsidy policies. Platforms must establish driver income protection funds, allocating a percentage of their commissions to purchase accident insurance for drivers, thereby enhancing industry stability and driving the development of the ride-hailing sector.

3.

Alleviate Radiation Concerns Toward ERVs

The two models in this study reveal that drivers’ radiation anxiety significantly influences their choice and decision-making regarding electric ride-hailing vehicles. For every unit increase in driver anxiety, the probability of their acceptance level rising by one tier decreases by 20%. Those completely unconcerned about electric vehicle radiation have a market-exit probability only 0.189 times lower than the reference group. Therefore, enhancing drivers’ and the public’s proper understanding of electric vehicle radiation can substantially reduce market instability. Underdeveloped cities and relevant authorities could draw lessons from successful practices in Shenzhen [57], innovatively incorporating electric vehicle radiation education into driver qualification training programs and making it part of the examination scope. This multi-dimensional approach to alleviating drivers’ radiation anxiety will ultimately improve their acceptance of electric ride-hailing services and stabilize the market order.

7. Conclusions

To explore the factors influencing the acceptance of ERVs and the exit-decision mechanisms among drivers in underdeveloped cities, based on 630 valid questionnaires investigated in Zhangzhou, Fujian, China, the statistical model was constructed to analyze the factors affecting drivers’ acceptance of ERVs and their decision to exit from the market. The results showed that, at the level of acceptance, (1) Ethical norms emerge as the primary driver of ERV adoption. Drivers’ environmental awareness and positive attitudes toward new-energy transportation significantly enhance their willingness to adopt ERVs, underscoring the pivotal role of pro-environmental values in this context. (2) Demographic data reveal unique local characteristics. Full-time ride-hailing drivers demonstrate a higher level of acceptance than other groups; however, the technology adoption of local full-time drivers lags noticeably. This result highlights risk aversion in underdeveloped regions. (3) The effectiveness of policy incentives has gradually weakened. Although 40.3% of drivers expect vehicle purchase subsidies and 35.7% anticipate operation subsidies, neither purchase nor operation subsidies reached the conventional level of statistical significance, contradicting research findings from developed regions. This phenomenon reflects policy implementation deviations and insufficient sensitivity among drivers in underdeveloped areas. (4) Beyond policy, user-perception traits diverge markedly by region. A notable finding is that while range anxiety constitutes a core barrier in developed cities, it shows no significant impact in this study. Instead, safety concerns—particularly worries about electromagnetic radiation—constitute a major psychological hurdle to ERV adoption. Drivers who feel more confident about radiation from vehicles are 1.199 times more likely to accept ERVs, and the interaction between social-media misinformation and health worries amplifies this psychological resistance.

As the key mechanisms underlying market-exit decisions, (1) Drivers’ acceptance of ERVs is positively and significantly correlated with their intention to remain in the market. Among those willing to stay, 87.83% show a high level of ERV acceptance, and this phenomenon is particularly prominent among full-time drivers. (2) Work status reveals a U-shaped relationship between work intensity and exit willingness, where both excessive and insufficient daily working hours trigger career turnover. This reflects the close connection between health issues, income concerns, and occupational stability. (3) Prior vehicle usage experience markedly shapes exit decisions. Drivers with no experience in using electric vehicles are far more likely to leave. This indicates that unfamiliarity and distrust of electric-vehicle technology constitute a critical barrier to the sustainable development of the ride-hailing industry. (4) Adjustments to platform revenue-sharing models, reductions in subsidies, and—most notably—the difficulty in obtaining dual certificates are regarded as the main external factors driving drivers to exit the market. Drivers who deem the dual certificate process “very difficult” exhibit a significantly lower retention rate than other drivers.

Finally, drawing on model results and the proven practices of developed cities, policy suggestions for underdeveloped regions were proposed, such as lowering dual-licensing thresholds, optimizing platform revenue-sharing models, enhancing health insurance coverage, and promoting electric vehicle education to stabilize ride-hailing capacity in underdeveloped cities.

This study has certain limitations. First, we only collected 630 questionnaires from a cross-sectional survey conducted in Zhangzhou City in November 2024. While representative, the dynamic effects of policy iterations and infrastructure improvements could be better analyzed with additional data. Second, while focusing on personal attributes, ethical norms, vehicle characteristics, and environmental factors, vehicle attributes are geographically influenced. For instance, battery performance is closely related to temperature [58,59]. Future research should explore regional disparities in underdeveloped cities, analyzing how geographical and economic development levels affect drivers’ acceptance and exit decisions to inform region-specific policies. Additionally, although existing studies have demonstrated that peer pressure and family factors significantly influence decision-making [60], this paper has not examined their marginal effects on electric ride-hailing adoption. Further investigation into social influences on acceptance and market-exit behaviors would provide valuable insights.

As for future research, we propose the following recommendations. First, researchers should align with policy release timelines to expand data samples while maintaining compliance. Second, we recommend examining how geographical environments moderate technology adoption by considering spatial heterogeneity such as climate variations, topography, and economic disparities in less-developed cities. Finally, future studies could adopt a mixed-method approach combining questionnaires and in-depth interviews, incorporating social network variables like peer effects to systematically evaluate how social factors influence drivers’ “acceptance–exit” trajectories.