Toward sustainable electric bus operations: a multidimensional structural model for future mobility management

1 Introduction

The transition toward green and low-carbon mobility has become a global strategic priority due to increasing greenhouse gas emissions, dependence on fossil fuels, and growing challenges in urban sustainability. In urban mobility, electric buses have emerged as an important component of sustainable public transportation systems due to their capability in reducing pollutant emissions, improving air quality, lower energy costs, and support low-carbon mobility and sustainability goals (Li et al., 2025; Ginting et al., 2024; Szczurowski et al., 2022; Yan et al., 2020; Xylia et al., 2019; Rogge et al., 2018; Song et al., 2018; He et al., 2018). However, their benefits still depend on environmental and economic performance, strongly influenced by operational efficiency in energy management, service schedule, resource utilization. In particular, unscheduled electric bus services, such as charter buses, tourism fleets, and demand-responsive transport services, often face operational challenges on fleet allocation, volatility in demand and service scheduling, charging infrastructure availability, and cost control. These challenges may reduce service effectiveness, limit vehicle utilization, and weaken the economic viability of electric bus operations despite their advantages.

Previous studies on electric buses have mainly focused on technical efficiency and operational optimization, including vehicle performance, battery technology, energy consumption, charging infrastructure planning, fleet size optimization, vehicle scheduling, and charging management. Several studies have also applied Life Cycle Assessment (LCA), well-to-wheel analysis, scenario analysis, and optimization models to evaluate environmental impacts, energy efficiency, and cost reduction in electric bus systems (Ginting et al., 2024; Szczurowski et al., 2022; Yan et al., 2020; Xylia et al., 2019; Rogge et al., 2018; Song et al., 2018; He et al., 2018; Gkiotsalitis et al., 2025; Zhou et al., 2023; Hu et al., 2022; Zhou et al., 2022; Liu et al., 2021; An, 2020; Lin et al., 2019; Ke et al., 2016; Jia and An, 2025; Zhou et al., 2024; Nath et al., 2024; Zeng et al., 2024; He et al., 2023; Gairola and Nezamuddin, 2023; Guschinsky et al., 2023; Dirks et al., 2022; Lee et al., 2021; He et al., 2020; Wang et al., 2017; Zeng et al., 2022; Zhang et al., 2021; Nordelöf et al., 2019; Fiori et al., 2021). These studies provide important insights into technical feasibility and environmental benefits of electric buses, particularly in scheduled public transportation. More recent research has also incorporated operational uncertainties, such as traffic variability, real-world usage, battery degradation, and non-linear charging characteristics, to improve the accuracy of electric bus planning under practical operating conditions (He et al., 2023; Zhang et al., 2021; Nordelöf et al., 2019).

In addition to fixed-route electric bus operations, a growing body of research has examined flexible and demand-responsive transport systems, which are more closely aligned with unscheduled electric bus services. Demand-responsive transport (DRT), flexible bus services, charter operations, and tourism-oriented fleets differ from conventional scheduled transit because their service patterns are influenced by dispersed demand, uncertain passenger volumes, varied departure times, route variability, and fluctuating dwell times. These characteristics additionally create operational complexity in fleet allocation, charging coordination, passenger assignment, and operational cost control. Recent studies on electric demand-responsive transport have highlighted the need to jointly consider vehicle dispatching, charging scheduling, fleet management, charging infrastructure availability, and demand uncertainty when designing sustainable electric mobility services. Similarly, studies on flexible bus and sightseeing bus operations emphasize the importance of adaptive scheduling, route planning, tourist assignment, and trade-offs between service quality and operational cost under variable demand conditions.

Demand-responsive transportation has been regarded as a mobility service characterized by flexible routes, service adaptability, and user-responsive operations, thereby providing a strong conceptual foundation for unscheduled transport systems (Deka et al., 2023). To support real-world operation, commercial sustainability of the transport has also been examined in relation to routing strategy, waiting time, travel time, service consistency, fleet utilization, and operational viability (Deka and Dilip, 2025). In addition, unmet DRT demand has also been analyzed to identify accessibility gaps and improve service coverage, which is directly relevant to passenger demand volatility (Al-Nabulsi et al., 2026). From a tourism aspect, previous studies have emphasized the role of shuttle services, direct tourist routes, and dynamic scheduling during peak tourist periods (Liu et al., 2025), as well as mobility challenges in transportation efficiency, scheduling, stakeholder coordination, and sustainable urban transitions (Tran et al., 2025). In summary, these studies indicate that unscheduled, demand-responsive, and tourism-oriented transport services require integrated consideration on flexible scheduling, uncertainty in passenger demand, route variability, fleet coordination, service quality, and operational sustainability.

Although previous studies provide useful insights into electric bus planning and optimization, the reviewed literature remains limited in its treatment of unscheduled and demand-responsive electric bus operations. Most electric bus models are based on fixed-route and timetable-based assumptions in which routes, stops, departure frequencies, operating periods, and charging opportunities are known in advance. These models generally optimize a specific decision, such as charger location, fleet size, vehicle scheduling, energy consumption, or charging time under predefined operating constraints (Perumal et al., 2022). They are effective for solving bounded planning problems but do not assess whether an operator has the combined technological, organizational, financial, and sustainability capabilities needed to manage services whose operating conditions change between assignments.

Research on demand-responsive transport (DRT) and flexible mobility captures some forms of operational variability, particularly dynamic routing, dispatching, passenger assignment, and real-time vehicle allocation. Its main analytical focus is often the matching of passengers, vehicles, and routes. Electric-vehicle-specific constraints, such as battery capacity, remaining range, charging duration, electricity cost, charger availability, and battery degradation, are not consistently integrated with service quality, organizational coordination, operating costs, and environmental management within the same framework. As a result, DRT optimization models may determine how vehicles should be routed or assigned under specified conditions, but they do not necessarily assess the broader operating capabilities required to sustain electric services over time.

This limitation is more pronounced for charter services and tourism fleets. These operations face irregular bookings, seasonal demand, changing destinations, variable trip distances and passenger loads, extended waiting periods, and uncertain access to charging facilities outside the operating base. Vehicle utilization depends on coordination among drivers, operation control centers, service staff, fleet managers, customers, and charging providers. Economic performance is affected by electricity consumption, overtime, maintenance, depreciation, vehicle downtime, and costs per operating hour or kilometer. Sustainability performance also depends on battery management, eco-driving practices, clean-energy use, emissions monitoring, and organizational reporting. The reviewed literature does not yet provide a validated system-level measurement framework that brings these interdependent conditions together for charter, tourism-fleet, and demand-responsive electric bus services. Table 1 summarizes these research streams and identifies the specific gap addressed by this study.

Table 1. Literature review matrix.

EBCP, Electric Bus Capability and Performance; SSUP, Service Scheduling and Utilization; OCE, Operational Cost Efficiency; SUO, Sustainability of Utilization and Operation; DRT, Demand-Responsive Transport. Full terms are used in the table to improve readability and reduce repeated cross-referencing.

To address this gap, the study develops and validates the Sustainable Electric Vehicle BUS Model (SEVBUS) as a higher-order measurement architecture for unscheduled electric bus operations. Its principal contribution is methodological: technological capability, service planning and utilization, cost efficiency, and sustainability-oriented operations are operationalized as a coherent structure of 48 observable indicators and 19 first-order factors under four dimensions: Electric Bus Capability and Performance (EBCP), Service Scheduling and Utilization Planning (SSUP), Operational Cost Efficiency (OCE), and Sustainability of Utilization and Operation (SUO). The framework was developed through structured literature synthesis, indicator screening, item-objective congruence assessment, Delphi consensus, exploratory analysis, and higher-order confirmatory measurement modeling. The resulting structure identifies operational domains that are empirically related but distinguishable within the studied context. It does not establish directional or causal relationships among the four dimensions.

While previous studies have primarily focused on optimizing individual operational decisions, such as vehicle routing, charging scheduling, fleet allocation, or energy management, SEVBUS addresses a different research objective by providing a validated measurement architecture for examining multidimensional operational conditions in unscheduled electric bus services. The framework integrates technological capability, service scheduling and utilization, operational cost efficiency, and sustainability within a single measurement structure. It does not generate optimized routing or charging solutions, nor does the present study establish that one domain causes changes in service performance or another domain. Within the studied context, the framework may provide a structured basis for future assessment of operational strengths and improvement opportunities before task-specific optimization methods are selected or implemented. Such diagnostic and planning uses remain proposed applications until the framework is evaluated with objective operational data.

The model has potential diagnostic and planning uses, but the present study validates its measurement structure rather than its effectiveness as a field-tested decision-support tool. Within the studied context, operators could use dimension-level and factor-level results to structure reviews of fleet and battery readiness, charging capability, schedule flexibility, dwell-time control, passenger-load balancing, overtime, maintenance, electricity costs, battery governance, eco-driving, and environmental reporting. Policymakers could use the same domains to examine operator data requirements, procurement criteria, sustainability reporting, and coordination between transport and energy agencies. These applications remain proposed uses until SEVBUS is implemented with actual operator data and tested against objective outcomes such as vehicle availability, energy consumption per kilometer, load factor, operating cost, and emissions performance.

Electric Bus Capability and Performance (EBCP): the dimension reflects technological and infrastructural limitations and efficiencies of electric buses that affect service flexibility.

Service Scheduling and Utilization Planning (SSUP): the dimension represents service scheduling management and bus utilization under charging constraints and flexible service conditions. This dimension is conceptually related to Demand-Responsive Transport (DRT) systems, which, while promoting accessibility and environmentally friendly transportation, continue to face challenges in cost–benefit viability and therefore require systematic operational management.

Operational Cost Efficiency (OCE): the dimension captures improvements in overall cost from coordinated decisions among fleet management, charging scheduling, charging duration and other systemic variables that contribute to volatility in the Total Cost of Ownership (TCO) of electric vehicles.

Sustainability of Utilization and Operation (SUO): the dimension operationalizes sustainability outcomes within a multidimensional conceptual framework. In this study, the environmental performance of electric buses is assumed to be empirically measurable (e.g., through Life Cycle Assessment: LCA) and highly sensitive to energy sources and actual operating conditions; therefore, it requires a clear operational management mechanism.

Transparent and auditable selection of factors and indicators was developed through a comprehensive literature review (Page et al., 2021). The factors were then assessed for relevance and appropriateness through Delphi expert consensus involving specialists in electric transportation, energy, and fleet management. The Delphi approach, with established guidance on expert selection, criteria development, and systematic reporting, is suitable for iterative screening of expert opinions (Okoli and Pawlowski, 2004). Following model development, SEM was used to evaluate the higher-order measurement structure, model fit, and pairwise correlations and covariances among the four latent dimensions. No directional paths among EBCP, SSUP, OCE, and SUO were estimated. Reliability and validity were evaluated using Average Variance Extracted (AVE), Composite Reliability (CR), and discriminant-validity criteria.

2 Research methodology

The development and assessment of structural equation modeling in electric bus operations require both theoretical rigor and methodological robustness. This necessity arises from the characteristics of the research phenomenon, which involves a multidimensional system encompassing automotive technical capacity, operational management, resource allocation, cost control and sustainability performance at both organizational and transportation system levels.

Although previous studies on electric mobility, fleet management, and sustainable transportation have emphasized specific dimensions such as energy efficiency, operational cost reduction, or carbon emissions, they often lack a comprehensive multidimensional measurement framework that integrates electric bus capability, operational management, cost efficiency, and sustainability performance. Therefore, this study adopts a sequential multi-stage research design to systematically develop, assess, and validate the SEVBUS measurement architecture. The research methodology employed in this study is based on three guiding principles.

• Content validity before measurement validation This principle confirms the appropriateness of factors and indicators before factor analysis and measurement-model validation. This step helps prevent empirical testing of a framework that lacks consistency with actual operational practices.
• Iterative expert consensus for construct refinement This principle utilizes the Delphi technique to reduce potential bias and enhance comprehensiveness of indicators through iterative consensus from the experts in relevant fields.
• Theory-driven measurement validation and association assessment via SEM This principle applies SEM to evaluate the higher-order measurement model, model fit, and non-directional associations among latent dimensions, while explicitly avoiding directional or causal interpretation where no directional paths were estimated.

2.1 Delphi technique

The Delphi-based expert evaluation was employed to establish content validity and assess the relevance of candidate indicators to real-world unscheduled electric bus operations (Cafiso et al., 2013; Ma et al., 2011; Mbakwe et al., 2016). The evaluation consisted of two sequential stages: an IOC-based content-validity assessment of the 48 indicators retained after preliminary screening of the 96-item literature-derived pool, followed by an expert consensus assessment using a structured five-point Likert-scale questionnaire. The same purposively selected panel of 13 experts participated in both stages. Using the same panel throughout both evaluation stages ensured consistency in expert judgments during the content validity assessment and the subsequent Delphi consensus process. Indicators satisfying the predefined content-validity and consensus criteria formed the final measurement set of 48 observed indicators, operationalizing 19 factors across the four dimensions of the SEVBUS framework. These indicators were subsequently used for survey instrument development, factor analysis, and higher-order measurement-model validation.

2.1.1 Expert selection and panel formation

The expert panel was formed through purposive sampling. Selection was based on the relevance of each expert’s qualifications and professional experience to the research topic, together with their ability to provide informed judgments on the candidate indicators. Thirteen experts meeting these criteria were recruited (Pumpugsri et al., 2023). Their areas of expertise included electric transportation, energy management, fleet management, and electric bus operations. The panel comprised representatives from academia, public agencies responsible for transportation and energy, and private organizations providing electric vehicle transportation services. The expert panel was intentionally designed to incorporate complementary perspectives from research, policy, energy management, and transport operations. Collectively, the experts possessed substantial professional experience in electric transportation, fleet management, transportation planning, and sustainable mobility, enabling the evaluation of the proposed indicators from both theoretical and practical perspectives. The diversity of the panel was intended to enhance the credibility and applicability of the framework within the studied context. This cross-sector composition incorporated theoretical, regulatory, and operational perspectives and reduced the likelihood that the evaluation reflected the views of a single stakeholder group.

2.1.2 Iterative Delphi evaluation and consensus assessment

The literature review yielded an initial pool of 96 candidate indicators covering the relevant conceptual domains of sustainable unscheduled electric bus operations. The number of indicators was determined by the literature identification and extraction process rather than by a predetermined numerical target. During preliminary screening, duplicate indicators were removed, conceptually overlapping indicators were consolidated, and indicators that were not applicable to unscheduled electric bus operations were excluded. This process reduced the initial pool from 96 to 48 indicators. The screening process aimed to retain indicators that were directly relevant to unscheduled electric bus operations while reducing conceptual redundancy. Duplicate or overlapping indicators were consolidated, and indicators specific to fixed-route services or optimization applications without direct operational relevance were excluded. The retained indicators collectively represent the technological, operational, economic, and sustainability dimensions required for subsequent expert evaluation and empirical validation. The 48 screened indicators were subsequently submitted to the purposively selected panel of 13 experts for content-validity assessment. The experts evaluated the congruence of each indicator with the research objectives and its intended SEVBUS factor using a three-category scale: −1 = incongruent, 0 = uncertain, and +1 = congruent. The Index of Item–Objective Congruence (IOC) was calculated for each indicator. All 48 indicators achieved IOC values of at least 0.50 and were therefore retained for the subsequent Delphi consensus evaluation.

A closed-ended questionnaire containing the 48 indicators was then administered to the same expert panel. The experts assessed the importance of each indicator using a five-point Likert scale ranging from 1 (lowest) to 5 (highest) (Ritmak et al., 2023). Item-level consensus required a median of at least 4, an interquartile range of no more than 1, and a standard deviation below 1. Kendall’s coefficient of concordance (W) was used to evaluate the overall level of agreement among the experts, with statistical significance assessed at p < 0.05. All 48 indicators satisfied the predefined item-level consensus criteria. The Delphi evaluation therefore confirmed a final set of 48 observed indicators, organized into 19 factors across the four dimensions of the SEVBUS framework. Although the Delphi process does not establish the completeness of the framework, it confirms that the retained indicators are relevant and appropriate for assessing sustainable unscheduled electric bus operations within the studied context.

2.2 Structural equation modeling (SEM)

Structural Equation Modeling (SEM) provides a framework for evaluating relationships between latent constructs and their observed indicators and, where explicitly specified, structural paths among latent variables (Madushika and Lu, 2025). In the present study, SEM was used primarily for higher-order measurement-model validation and for estimating associations among the four SEVBUS dimensions. EFA was conducted using SPSS for preliminary assessment of the empirical grouping of the literature- and Delphi-derived indicators, followed by CFA and higher-order SEM using AMOS. The analysis evaluated factor loadings, model fit, reliability, convergent validity, discriminant validity, and correlations and covariances among EBCP, SSUP, OCE, and SUO. Because no directional paths among these four dimensions were estimated, the inter-construct findings are interpreted as associations rather than directional or causal effects.

2.2.1 SEM procedure for the SEVBUS model

Following the Delphi process and confirmation of content validity, the observed variables were consolidated into 48 indicators categorized into four dimensions of the SEVBUS model: Electric Bus Capability and Performance (EBCP), Service Scheduling and Utilization Planning (SSUP), Operational Cost Efficiency (OCE), and Sustainability of Utilization and Operation (SUO), then assessed for its appropriateness by using the Kaiser–Meyer–Olkin (KMO) Measure of Sampling Adequacy, with an acceptance criterion of 0.70 or higher and Bartlett’s Test of Sphericity (Ritmak et al., 2023).

2.2.2 Factor analysis and model fit assessment

Once the data were confirmed to be suitable for analysis, Exploratory Factor Analysis (EFA) was conducted to examine the empirical grouping and underlying factor structure of the indicators. Confirmatory Factor Analysis (CFA) was then performed to evaluate the proposed relationships between the latent constructs and their observed indicators and to assess the consistency of the measurement model with the empirical data using two categories of fit indices:

• Absolute fit indices include CMIN/DF, RMSEA, GFI, AGFI and RMR.
• Incremental fit indices include NFI, TLI, CFI and IFI

Following assessment of measurement-model fit, the correlations and covariances among the four latent SEVBUS dimensions were examined to evaluate the proposed non-directional hypotheses of association.

The study employed Structural Equation Modeling (SEM) because it enables the simultaneous evaluation of multiple latent constructs and their observed indicators within a unified analytical framework. In the present study, SEM was applied to assess the higher-order measurement architecture, model fit, measurement reliability and validity, and the associations among the four SEVBUS dimensions. The analysis was not designed to estimate directional paths or causal effects among EBCP, SSUP, OCE, and SUO. Accordingly, the inter-construct findings are interpreted as statistical associations within the proposed multidimensional framework.

2.2.3 Measurement validity and reliability

Measurement reliability and validity were evaluated using Composite Reliability (CR) and Average Variance Extracted (AVE). CR values were required to exceed 0.70 and AVE ≥ 0.50. Convergent validity was supported when AVE ≥ 0.50. Discriminant validity was assessed by comparing AVE with Maximum Shared Variance (MSV), where MSV should be lower than AVE to support the distinctiveness of the latent constructs. These analyses evaluate the adequacy of the SEVBUS measurement framework. Associations among the four latent dimensions were subsequently examined using correlations and covariances and were interpreted as non-directional relationships (Figure 1).

3 Results

3.1 Reliability of the Delphi technique

The result analysis shown in Table 2 demonstrates that the 48 indicators under the SEVBUS model, classified into four main dimensions: Electric Bus Capability and Performance (EBCP), Service Scheduling and Utilization Planning (SSUP), Operational Cost Efficiency (OCE) and Sustainability of Utilization and Operation (SUO), are consistent with experts’ opinion at a high level. The Delphi technique results indicate that the IQR is 0.41, with a median of 4.80 and a standard deviation of 0.36. These values fall within the acceptance criteria and show a strong level of consensus. Moreover, Kendall’s Coefficient of Concordance (W = 0.438, p < 0.001, N = 13) indicates moderate to high agreement among experts. Thus, affirming the appropriateness and reliability of the 48 indicators. Therefore, the indicators were considered suitable for subsequent factor analysis and higher-order measurement-model validation.

Table 2. Delphi results.

IQR, interquartile range; M. D., median; AVR., average; S. D., standard deviation; KW, Kendall’s Coefficient of Concordance; Level of significance; N, number of experts.

3.2 The SEVBUS model

Following empirical validation, the SEVBUS model comprises four dimensions and nineteen factors representing technological capability, service planning and utilization, operational cost efficiency, and sustainability-related operations in unscheduled electric bus services.

3.2.1 Electric bus capability and performance dimension (EBCP)

• This dimension reflects the technical readiness and operational robustness of electric buses in real service environments. It includes four factors: Average Operating Speed (AOS), Vehicle Lifespan and Performance (VLP), Electric Energy Performance Efficiency (EE) and Battery Capacity and Charging Capability (BCC). Average Operating Speed (AOS): this factor demonstrates the capability of electric buses to maintain a stable speed under varying traffic conditions. It consists of three indicators: Average Operating Speed (AOS1), Speed Stability across Traffic Conditions (AOS2), and Speed Consistency across Routes and Time Periods (AOS3). The high expert ratings indicate that stable operating speed was regarded as a relevant technical characteristic of the AOS factor and of service reliability within the EBCP construct (Sun and Chen, 2024).
• Vehicle Lifespan and Performance (VLP): this factor addresses the durability and reliability of electric buses during their operational life. It comprises three indicators: Service Life and Operational Performance (VLP1), Vehicle Reliability and Uptime over Lifecycle (VLP2), and Performance Degradation Control over Time (VLP3). The high expert ratings indicate that vehicle lifespan, reliability, and performance stability were regarded as relevant characteristics of the VLP factor.
• Electric Energy Performance Efficiency (EE): this factor focuses on the efficiency of energy consumption. It is measured through two indicators: Energy-efficiency Range per Charge (EE1) and Energy Consumption Index, measured in kWh/km (EE2). Strong consistency was measured from experts’ opinions, indicating that energy efficiency plays a key role in both economic viability and environmental impact reduction, as it lowers the cost per distance and reduces energy consumption (Li et al., 2025).
• Battery Capacity and Charging Capability (BCC): this factor reflects the adequacy of battery systems and charging infrastructure, assessed through two indicators: Battery Energy Capacity (BCC1) and Smart Charging Power (BCC2). The high level of consensus emphasizes strong impacts from battery capacity and continuous charging capability on operational stability and service continuity as they allow the operation to perform smoothly.

3.2.2 Service scheduling and utilization planning dimension (SSUP)

• This dimension focuses on planning efficiency, demand management and service quality. It includes six factors: Trip Start Consistency (TSC), Average Visiting Duration (AVD), Passenger Capacity Utilization (PCU), Passenger Load per Trip (PLT), Company Performance Efficiency (CPE)and Passenger Satisfaction Index (PSI). Trip Start Consistency (TSC): this factor evaluates the punctuality and flexibility of bus scheduling through two indicators: Trip Start Time (TSC1) and Flexible Scheduling Time (TSC2). High scores in median and average with low standard deviation indicate that a consistent schedule with operational flexibility is essential for effective unscheduled electric bus services, as the operator can forecast and manage an operation and prepare to serve unexpected demands.
• Average Visiting Duration (AVD): this factor reflects management efficiency in controlling passenger boarding and alighting time. It includes three indicators: Average Visiting Duration [Planning Perspective (AVD1)]; Adherence to Planned Dwell Time (AVD2); and Control of Dwell Time Variability (AVD3). The high scores from the experts emphasize that effective control of visiting duration is important for improving operational efficiency due to its ability to oversee the overall operation.
• Passenger Capacity Utilization – Load Factor (PCU): this factor assesses the effectiveness of capacity utilization through three indicators: Passenger Capacity Utilization (PCU1); Seat Occupancy Rate during Operation (PCU2); and Optimization of Peak-load Utilization (PCU3). High consensus among experts confirms that optimizing load factor is critical for sustainable service operations, as it allows the bus operators to serve all passengers regardless of fluctuating demand.
• Passenger Load per Trip (PLT): this factor focuses on demand distribution efficiency through three indicators: Passenger Load per Trip (PLT1); Load Consistency across Similar Trips (PLT2); and Passenger Load Balancing between Trips (PLT3). The high expert ratings indicate that passenger-load allocation and balancing were regarded as relevant components of service utilization within the PLT factor.
• Company Performance Efficiency (CPE): this factor reflects effectiveness in organizational management and operational coordination from three indicators: Company Management Efficiency (CPE1); Coordination Effectiveness among Operations (CPE2); and Responsiveness in Operational Decision-making (CPE3). The high expert consensus indicates that management efficiency, operational coordination, and responsiveness were regarded as relevant components of the CPE factor.
• Passenger Satisfaction Index (PSI): this factor reflects service quality from the user perspective, measured through three indicators: Passenger Satisfaction Index (PSI1); Satisfaction with Service Reliability and Punctuality (PSI2); and Satisfaction with Comfort and Service Experience (PSI3). The expert ratings indicate that passenger satisfaction, service reliability, punctuality, and comfort were regarded as relevant service-quality components within the PSI factor.

3.2.3 Operational cost efficiency dimension (OCE)

This dimension reflects financial sustainability. It includes four factors: Maintenance & Depreciation Cost (MDC), Labor & Time Cost (LTC), Electricity Cost Efficiency (ECE)and Total Operation Cost (TOC).

• Maintenance & Depreciation Cost (MDC): this factor refers to equipment expenses in the operation, measured through two indicators: Maintenance Cost (MDC1) and Vehicle Depreciation Cost (MDC2). The high level of consensus reflects that maintenance and depreciation, which are relatively high in capital intensive industry like bus operation, are major components in operational cost and require attentive management across the vehicle life cycle.
• Labor & Time Cost (LTC): as the operation involves manpower, this factor evaluates labor-related costs through two indicators: OT Management Cost (LTC1) and Driver Overtime Cost (LTC2). A high consensus from the experts indicates that manpower management is highly related to efficiency in cost allocation.
• Electricity Cost Efficiency (ECE): this factor considers utility expenses as a major cost, as electric buses consume electricity rather than fossil fuels. The factor measures the efficiency of electricity expenditure through two indicators: Electricity Cost per Main Operation (ECE1) and Electricity Cost per Extended Service (ECE2). A high score in the consensus shows that electricity cost largely contributes to expenses and requires strong attention for operational efficiency.
• Total Operation Cost (TOC): this factor reflects overall cost management efficiency through two indicators: Service Cost per Operation Hour (TOC1) and Total Cost per Km Index (TOC2). The experts’ opinions imply that operational cost is inevitably related to efficiency in bus operation, with full consideration in terms of operation time and distance required.

3.2.4 Sustainability of utilization and operation dimension (SUO)

This dimension emphasizes environmental performance and responsible resource utilization. It includes five factors: Resource & Green Energy Efficiency (RGEE), Battery Energy and Waste Management (BEWM), Eco-Safe Driving Behavior (ESDB), CO₂ Emission Reduction per Trip (CERT) and Green Operation Index (GOI).

• Resource & Green Energy Efficiency (RGEE): this factor refers to utilization of green energy in operation, measured through two indicators: Resource Utilization Efficiency (RGEE1) and Efficient Use of Clean and Renewable Energy (RGEE2). The high score from experts supports the concept of low-carbon transportation and clean energy operation.
• Battery Energy and Waste Management (BEWM): this factor evaluates environmental risk management on battery systems through two indicators: Battery Waste Management (BEWM1) and Battery Lifecycle Monitoring and Disposal Control (BEWM2). A strong consensus implies that strong frameworks and operational guidelines for battery management in the varied life cycle are necessary for operational sustainability.
• Eco-Safe Driving Behavior (ESDB): this factor reflects environmentally responsible driving behaviors from three indicators: Eco-safe Driving Behavior (ESDB1); Smooth Driving and Reduced Harsh Maneuvers (ESDB2); and Compliance with Eco-Driving Guidelines (ESDB3). A high-level of experts’ consensus indicates that eco-safe driving behaviors contribute to sustainable operation by lowering energy consumption, reducing accidents and creating environmental awareness in daily life.
• CO₂ Emission Reduction per Trip (CERT): this factor measures the reduction of environmental impact from bus operation through three indicators: CO₂ Emission Reduction per Trip (CERT1); Consistency of CO₂ Reduction across Routes (CERT2); and CO₂ Reduction Compared with Conventional Buses (CERT3). The experts’ consensus shows a large expectation for the reduction of CO₂ emissions to eco-friendly operation.
• Green Operation Index (GOI): this factor reflects organizational commitment to sustainable operation from three indicators: Green Operation Index (GOI1); Implementation of Green Operational Practices (GOI2); and Monitoring and Reporting of Green Operation (GOI3). The high expert ratings indicate that green operational practices, monitoring, and reporting were regarded as relevant components of the GOI factor within the SUO construct.

3.3 Measurement reliability and data adequacy assessment

Structured questionnaires were distributed to 400 respondents involved in unscheduled electric bus services. The respondents were selected by purposive criterion-based sampling methods due to the requirement on sufficient knowledge, understanding, and operational experience as well as applicable license in related areas such as bus driving, fleet management, and operational management. The respondents included electric bus drivers, operation control center staff, bus stop or station personnel, and operational managers, all of whom play important roles in planning, controlling, and performing day-to-day operations with direct experience on the efficiency and sustainability of electric bus services.

The final sample consisted of electric bus drivers (n = 280, 70.00%), operation control center staff (n = 50, 12.50%), bus stop/station/service staff (n = 50, 12.50%), and operational/fleet managers (n = 20, 5.00%), reflecting operational structure of unscheduled electric bus services with regards to frontline operational experience and managerial perspectives.

Following data collection, the study assessed internal consistency with Cronbach’s alpha of 0.95 exceeds the acceptable threshold of 0.70, indicating high internal consistency. Furthermore, the adequacy of the data for factor analysis was examined using the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s Test of Sphericity. The analysis indicated a KMO value of 0.925, which is above the acceptance criteria of 0.50 and represents an excellent level of sampling adequacy. Moreover, Bartlett’s Test of Sphericity was statistically significant (p < 0.001), indicating that the correlation matrix differs significantly from an identity matrix and is appropriate for factor analysis.

3.4 Exploratory factor analysis

Exploratory Factor Analysis (EFA) was then conducted to examine the empirical structure of the 19 factors and 48 observed indicators within the SEVBUS model to assess whether the observed indicators, derived from the literature review and validated through expert consensus, demonstrated an empirically coherent structure. The results were interpreted as supportive evidence for the grouping of indicators within the SEVBUS framework.

Based on consistency and expert evaluations, the authors recategorized the indicators shown in Table 2 before conducting the factor analysis and extracted factors using Principal Component Analysis (PCA) by selecting factors with eigenvalues greater than 1, according to Kaiser’s criterion. The factors were then rotated using varimax rotation to enhance the clarity of interpretation. The results show that all indicators had factor loadings and communalities greater than 0.50, indicating their suitability for factor analysis and consistency with the proposed recategorization, as shown in Table 2.

Based on expert recommendations and the empirical results from the EFA, the 19 factors were grouped into four dimensions: EBCP, SSUP, OCE, and SUO, representing technical capability, service performance, sustainability outcomes, and cost efficiency in electric bus operations. Subsequently, all factors were tested by Confirmatory Factor Analysis (CFA), with the results shown in Table 3.

Table 3. CFA factor loading statistics.

S. E., standard error; C.r., critical ratio; unstandardized. p < 0.001 for all coefficients (***).

3.5 Confirmatory factor analysis

After the exploratory factor analysis (EFA), Confirmatory Factor Analysis (CFA) was conducted using AMOS to evaluate the proposed measurement structure of the SEVBUS model. The CFA results show that the observed variables have statistically significant factor loadings, providing empirical support for their representation of the corresponding latent constructs identified through the EFA.

The assessment of model fit by absolute fit indices found that the ratio of chi-square to degrees of freedom (CMIN/DF) is 1.689, below the recommended threshold of 3.00, suggesting a good level of model fit. Furthermore, the Root Mean Square Error of Approximation (RMSEA) is 0.041, below 0.05, indicating an excellent level of fit between the proposed model and the empirical data. In addition, the Goodness-of-Fit Index (GFI) and Adjusted Goodness-of-Fit Index (AGFI) are 0.951 and 0.924, exceeding the acceptance criteria (GFI > 0.95 and AGFI > 0.90). Also, the Root Mean Square Residual (RMR) is 0.022, which is close to zero, indicating a low level of residual variance.

In terms of incremental fit indices, the results indicate that the SEVBUS model demonstrates an excellent level of fitness with the Normed Fit Index (NFI) of 0.976, the Tucker–Lewis Index (TLI) of 0.986, the Comparative Fit Index (CFI) of 0.990 and the Incremental Fit Index (IFI) of 0.990. All these values exceed the recommended value of 0.95, indicating that the proposed model demonstrates substantially better fit than the independence model. Although the chi-square statistic at 121 degrees of freedom is 204.361 (χ² = 204.361, p = 0.000), which may indicate a lack of model fit, it should be noted that the chi-square statistic is highly sensitive to sample size, particularly in studies with relatively large samples, as in this research. Accordingly, the overall pattern of fit indices supports an acceptable-to-good fit between the proposed measurement model and the empirical data.

3.6 Measurement validity and reliability of the SEVBUS model

Table 4 presents the analysis results of Average Variance Extracted (AVE) and Composite Reliability (CR) derived from the CFA of the SEVBUS model at both the dimensional level and factor level within each dimension. The results show that the AVE and CR values for all dimensions and factors are at a satisfactory level, meeting the acceptance criteria of AVE ≥ 0.50 and CR ≥ 0.70. Therefore, the results support the reliability and convergent validity of the SEVBUS measurement model. The standardized factor loadings (λᵢ or Li) were subsequently examined to assess the empirical representation of the higher-order construct and its dimensions.

Table 4. The SEVBUS model.

The standardized factor loading (L i); variance (L i 2); the error variance 1 − L i 2(e i); C. R., Composite (construct) Reliability; AVE., Average Variance Extracted; C. A., Cronbach’s Alpha; Acceptable (*).

The analysis at the dimensional level indicates that the Sustainability of Utilization and Operation (SUO) dimension has the highest factor loading, with Li = 0.83, indicating that the SUO dimension exhibits the strongest representation of the higher-order SEVBUS construct among the four dimensions. Within the SUO dimension, the Green Operation Index (GOI), CO₂, CO₂ Emission Reduction per Trip (CERT) and Battery Energy and Waste Management (BEWM) factors show the highest loadings, with Li values of 0.87, 0.76, and 0.76, respectively. These findings indicate that GOI, CERT, and BEWM provide strong empirical representation of the Sustainability of Utilization and Operation construct, collectively reflecting organizational, environmental, and battery management aspects.

The dimension with the second-highest factor loading is Operational Cost Efficiency (OCE), with Li = 0.77. Within this dimension, Total Operation Cost (TOC), Labor & Time Cost (LTC), Maintenance and Depreciation Cost (MDC) and Electricity Cost Efficiency (ECE) have Li values of 0.96, 0.80, 0.78, and 0.76, respectively. These standardized loadings indicate that TOC, LTC, MDC, and ECE adequately represent different components of the Operational Cost Efficiency construct, including capital-related costs, labor, maintenance, and electricity expenditure. Collectively, these indicators provide empirical support for the construct representation within the measurement model.

The third-highest factor loading is found in the Electric Bus Capability and Performance (EBCP) dimension, with Li = 0.67. Within this dimension, Average Operating Speed (AOS), Vehicle Lifespan and Performance (VLP), Electric Energy Performance Efficiency (EE) and Battery Capacity and Charging Capability (BCC) have Li values of 0.97, 0.88, 0.85, and 0.51, respectively. These results indicate that the selected indicators adequately represent different technical characteristics underlying the Electric Bus Capability and Performance construct. These indicators collectively represent complementary technical characteristics underlying the Electric Bus Capability and Performance construct.

Lastly, the Service Scheduling and Utilization Planning (SSUP) dimension has a factor loading of Li = 0.64, which remains within the acceptable level. Within this dimension, Passenger Load per Trip (PLT), Passenger Capacity Utilization – Load Factor (PCU), Trip Start Consistency (TSC), Average Visiting Duration (AVD), Company Performance Efficiency (CPE) and Passenger Satisfaction Index (PSI) exhibit Li values ranging from 0.93 to 0.98. These findings indicate that the selected indicators consistently represent multiple operational attributes associated with service scheduling and utilization planning. Collectively, these indicators characterize operational planning and service utilization within the SEVBUS measurement framework.

All measurement statistics are presented in Table 4, while the higher-order measurement configuration is illustrated in Figure 2. The results support the construct validity and reliability of the proposed SEVBUS measurement framework. Its practical application as a diagnostic or planning tool requires further validation through implementation in operational settings.

The higher-order measurement model illustrated in Figure 2 presents the proposed construct architecture of the Sustainable Electric Vehicle BUS Model (SEVBUS). The following hypotheses were formulated to assess model fit and the non-directional associations among the four SEVBUS dimensions:

H1: The proposed SEVBUS higher-order measurement model demonstrates an acceptable fit with the empirical data.

H2: Electric Bus Capability and Performance (EBCP) is significantly correlated with Operational Cost Efficiency (OCE).

H3: Operational Cost Efficiency (OCE) is significantly correlated with Service Scheduling and Utilization Planning (SSUP).

H4: Operational Cost Efficiency (OCE) is significantly correlated with Sustainability of Utilization and Operation (SUO).

H5: Electric Bus Capability and Performance (EBCP) is significantly correlated with Sustainability of Utilization and Operation (SUO).

H6: Service Scheduling and Utilization Planning (SSUP) is significantly correlated with Sustainability of Utilization and Operation (SUO).

H7: Electric Bus Capability and Performance (EBCP) is significantly correlated with Service Scheduling and Utilization Planning (SSUP).

The H1 hypothesis was tested by applying multiple indices to verify the consistency and confirm the reliability and accuracy of the analysis, as presented in Table 5. Hence, H1 is supported, indicating that the proposed higher-order SEVBUS measurement model demonstrates an acceptable level of fit with the empirical data. This finding concerns measurement-model fit and should not be interpreted as evidence of directional or causal relationships among the four SEVBUS dimensions, as shown in Figure 2. The higher-order relationship is represented by Equation (1), while the first-order measurement relationships are represented by Equations (2)–(5).

The SEVBUS = 0.67 EBCP eb + 0.83 SUO su + 0.64 SSUP ss + 0.77 OCE oc (1)

EBCP eb = 0.97 e r 1 + 0.88 e r 2 + 0.85 e r 3 + 0.51 e r 4 + ε eb (2)

ε eb = E er 1 + E er 2 + E er 3 + E er 4

SUO su = 0.75 e r 5 + 0.76 e r 6 + 0.70 e r 7 + 0.76 e r 8 + 0.87 e r 9 + ε su (3)

ε su = E er 5 + E er 6 + E er 7 + E er 8 + E er 9

SSUP ss = 0.95 e r 10 + 0.95 e r 11 + 0.96 e r 12 + 0.98 e r 13 + 0.93 e r 14 + 0.94 e r 15 + ε ss (4)

ε ss = E er 10 + E er 11 + E er 12 + E er 13 + E er 14 + E er 15

OCE oc = 0.78 e r 16 + 0.80 e r 17 + 0.76 e r 18 + 0.96 e r 19 + ε oc (5)

ε oc = E er 16 + E er 17 + E er 18 + E er 19

Table 5. Measurement of model fit.

Table 6 presents discriminant validity and the pairwise correlations among the four SEVBUS dimensions. All six correlations are statistically significant at p < 0.001. Accordingly, the results provide statistical support for H2-H7 as non-directional hypotheses of association: H2, EBCP-OCE (r = 0.54); H3, OCE-SSUP (r = 0.50); H4, OCE-SUO (r = 0.61); H5, EBCP-SUO (r = 0.56); H6, SSUP-SUO (r = 0.56); and H7, EBCP-SSUP (r = 0.38). These coefficients indicate statistical associations and do not establish directionality or causality. In addition, the Maximum Shared Variance (MSV) values are lower than the corresponding Average Variance Extracted (AVE) values, supporting discriminant validity among the four dimensions.

Table 6. Discriminant validity of interrelated dimensions in the SEVBUS model.

Cor., correlations; Cov., covariances; MSV, Maximum Shared Variance; S. E., standard error; C. R., critical ratio; unstandardized. p < 0.001 for all coefficients significant (***).

The correlations among the four latent dimensions, as illustrated in Figure 3, are statistically significant and support the interpretation of EBCP, SSUP, OCE, and SUO as related but empirically distinguishable domains within the SEVBUS measurement framework. These findings support the multidimensional measurement structure but do not establish directionality, causality, or operational impact.

The SEVBUS framework comprises four empirically interrelated dimensions: Electric Bus Capability and Performance (EBCP), Service Scheduling and Utilization Planning (SSUP), Sustainability of Utilization and Operation (SUO), and Operational Cost Efficiency (OCE). The statistically significant correlations indicate that these technical, operational, economic, and sustainability-related domains are associated within the studied sample while remaining empirically distinguishable. The estimated correlations do not establish the direction of influence or causal relationships among the dimensions.

Within the Electric Bus Capability and Performance (EBCP) dimension, factors such as Average Operating Speed, Vehicle Lifespan and Performance, Electric Energy Performance Efficiency and Battery Capacity and Charging Capability represent key technical characteristics of the Electric Bus Capability and Performance dimension. The EBCP dimension captures technical characteristics associated with vehicle capability, operational performance, and energy efficiency. The practical implications of these characteristics should be considered together with the observed inter-construct correlations and require further validation using objective operational data.

In the Service Scheduling and Utilization Planning (SSUP) dimension, Passenger Capacity Utilization, Load Factor, Company Performance Efficiency and Passenger Satisfaction Index serve as key indicators reflecting the social dimensions of sustainability in public transportation services. When considering each factor, The SSUP dimension represents operational characteristics associated with service scheduling, capacity utilization, organizational coordination, and passenger-related outcomes. Their operational implications should be interpreted in conjunction with the observed associations among the SEVBUS dimensions rather than as effects demonstrated by directional structural paths.

At the same time, the Sustainability of Utilization and Operation (SUO) dimension emphasizes environmentally responsible operations, including Resource & Green Energy Efficiency, Battery Energy and Waste Management, Eco-Safe Driving Behavior and CO₂ Emission Reduction per Trip. While vehicle capability and service operation reflect traditional business, this dimension incorporates sustainability-related factors. It emphasizes environmental responsibility through the use of clean energy, an end-to-end perspective on energy usage, consideration of driving behaviors for the safety of road users and passengers and inclusion of carbon emissions into perspective. These indicators collectively represent sustainability-related operational characteristics incorporated within the proposed framework. Their relationship with other operational domains should be interpreted on the basis of the observed correlations; their effects on operational performance were not directly tested.

Furthermore, the Operational Cost Efficiency (OCE) dimension includes Maintenance & Depreciation Cost, Labor & Time Cost and Electricity Cost Efficiency. The factors in this dimension operationalize the economic dimension of the proposed measurement framework. As they are key components in operating electric buses, representing economic dimension of sustainability in electric bus operations.

The SEVBUS model integrates technical, economic, service-related, and environmental domains within a validated measurement structure (Figure 4). The framework may provide a basis for future policy analysis, transportation planning, and operational assessment, but these applications require further field implementation and validation.

In addition to the factors studied, the environmental conditions of bus routes also play a significant role in the operational efficiency of smart electric bus systems, affecting service stability, energy consumption and system reliability. Road conditions, such as slope, surface conditions, traffic congestion and infrastructure limitations, directly influence driving patterns and resource deployment in electric bus operations. As a result, analyzing route environments provides a deeper understanding of the factors affecting the systematic efficiency of the SEVBUS framework and supports improvements in infrastructure planning for long-term operations.

Furthermore, route conditions are directly related to the Electric Bus Capability and Performance Dimension of electric buses on energy consumption, mobility steadiness and battery efficiency. For instance, steep slopes may increase energy consumption, while poor road surfaces or inadequate drainage systems may disrupt smooth fleet operations and reduce schedule stability. The analysis supports vehicle design and fleet planning with route conditions. At the same time, route conditions also relate to the Service Scheduling and Utilization Planning Dimension, since heavy traffic congestion or infrastructure constraints may affect fleet travel time, schedule punctuality, passenger load distribution and user satisfaction, which are key indicators within the social sustainability dimension.

Therefore, examining the relationship between route environments and service performance helps explain the dynamics of electric bus systems in real-world operations. Such analysis supports policy-level decision-making and sustainable public transportation management, ultimately contributing to the long-term effectiveness of electric bus operation systems.

4 Discussion

This study emphasizes that sustainability challenges in electric bus operations should not be viewed merely as routine operational concerns, but rather as systemic issues that simultaneously influence: (1) operational efficiency and resource utilization; (2) overall operating costs; and (3) sustainability performance, including energy efficiency, environmental outcomes, and service quality within tourism and demand-responsive transportation environments. These challenges are particularly significant in operational contexts characterized by fluctuating demand and dynamic service requirements.

The development and empirical validation of the Sustainable Electric Vehicle BUS Model (SEVBUS) provide a methodological contribution by establishing an integrated measurement framework for unscheduled electric bus operations. Existing studies frequently examine technological, operational, cost, or sustainability-related components separately. SEVBUS addresses this fragmentation by organizing these domains within a coherent higher-order measurement architecture. The present results establish empirical associations among the four dimensions but do not demonstrate directional mechanisms through which one dimension produces changes in another. Sphericity (p < 0.001) confirms that the dataset is statistically suitable for factor analysis.

4.1 Expert consensus and empirical validation

This study employed the Delphi technique to select relevant indicators and confirm their appropriateness through consensus from 13 experts to help reduce the risk of “theoretical fit but practical insignificance” often observed in conceptual model development. The Kendall’s coefficient of concordance (W = 0.438, p < 0.001) indicates a moderate-to-high level of consensus, which is considered appropriate for complex research problems requiring integrated perspectives such as fleet operation, energy and management. The quantitative data collected from 400 stakeholders, which is deemed to have strong reliability and sampling adequacy, were tested and resulted in the Cronbach’s alpha coefficient of 0.95, indicating a high internal consistency. Moreover, the Kaiser–Meyer–Olkin (KMO) measure of 0.925 and Bartlett’s Test of Sphericity (p < 0.001) confirm that the dataset is statistically suitable for factor analysis.

4.2 SEVBUS model fitness and supporting fit indices

The CFA and higher-order SEM results demonstrated acceptable model fit across the reported indices, supporting the empirical consistency of the SEVBUS measurement framework. Convergent validity, construct reliability, and discriminant validity indicate that the four dimensions are empirically supported and distinguishable. These measurement properties provide a basis for examining associations among the dimensions but do not demonstrate directional effects on sustainability performance or operational effectiveness.

4.3 Multi-dimensional interrelationships for sustainable electric bus operations

A key empirical finding is that all six pairwise correlations among EBCP, SSUP, SUO, and OCE are positive and statistically significant (r = 0.38–0.61, all p < 0.001). These results support interpretation of the four dimensions as interrelated operational domains within the SEVBUS framework. They do not demonstrate that sustainability outcomes depend on synergistic interactions among the dimensions or that the dimensions mutually reinforce one another, because directionality was not tested.

From policy and managerial perspectives, the observed associations suggest that technological capability, service planning, operational coordination, cost efficiency, and sustainability-related practices should be considered jointly when assessing unscheduled electric bus operations. However, the present findings do not demonstrate that investment in any single domain is sufficient or insufficient to produce sustainable transportation outcomes, nor do they establish that improvements in one domain lead to improvements in another. The results instead provide a multidimensional basis for future operational assessment and empirical investigation.

From an operational perspective, the SEVBUS framework may provide a basis for future evaluation as a structured diagnostic tool. Operators could potentially use the four dimensions to organize assessments of vehicle capability, service scheduling, operational costs, and sustainability practices. However, the present study did not test whether SEVBUS-based assessments improve operational decisions or outcomes. Its diagnostic and decision-support effectiveness should therefore be evaluated through field implementation using objective operational data.

4.4 Factor loading (Li)

At the dimensional level, the Sustainability of Utilization and Operation dimension has the highest factor loading (Li = 0.83), followed by Operational Cost Efficiency (OCE) with Li of 0.77, Electric Bus Capability and Performance (EBCP) with Li of 0.67 and Service Scheduling and Utilization Planning (SSUP) with Li of 0.64. The higher standardized loading indicates that the SUO dimension shares a larger proportion of variance with the higher-order SEVBUS construct. The high loadings of GOI, CERT, and BEWM indicate that these indicators consistently represent the SUO construct.

Similarly, the strong factor loadings observed within the Operational Cost Efficiency (OCE) dimension indicate that overall operational costs—including Total Operational Cost, Labor and Time Cost, Maintenance and Depreciation Cost, and Electricity Cost Efficiency—The high standardized loadings indicate that these indicators consistently represent multiple components of the Operational Cost Efficiency construct. Their managerial implications should be interpreted together with the observed inter-construct associations and not inferred from the measurement loadings alone.

From a measurement perspective, SSUP represents an operational construct within the proposed framework. This dimension encompasses Trip Start Consistency (TSC), Average Visiting Duration (AVD), Passenger Capacity Utilization (PCU), Passenger Load per Trip (PLT), Company Performance Efficiency (CPE), and Passenger Satisfaction Index (PSI). These factors represent scheduling, capacity utilization, organizational coordination, and passenger-related aspects of unscheduled electric bus operations. In tourism and demand-responsive settings, these attributes are conceptually relevant to operational coordination and service utilization. Their effects on service reliability, resource allocation, or sustainability outcomes were not directly tested in the present study.

4.5 Integrated user experience and operational efficiency in sustainable electric bus service

The inclusion of Passenger Satisfaction Index (PSI) and Company Performance Efficiency (CPE) broadens the SEVBUS measurement architecture by incorporating passenger- and organization-related attributes alongside technical, cost, and environmental factors. These attributes are relevant considerations in tourism and demand-responsive transportation, but implications for repeat usage, service loyalty, destination competitiveness, and broader sustainability outcomes are conceptual or literature-based rather than causal findings from the present analysis. Within SEVBUS, PSI and CPE should therefore be interpreted primarily as components of the validated measurement architecture.

4.6 Theoretical implication and future research

From a theoretical perspective, SEVBUS should be interpreted primarily as an integrated measurement framework rather than as an explanatory causal model. It organizes technological capability, service scheduling and utilization, operational cost efficiency, and sustainability-oriented operations within a coherent construct architecture and establishes empirical associations among these domains. The present study does not demonstrate a multidimensional causal mechanism or that sustainability is a result of co-management among the four dimensions. Directionality and causal mechanisms remain open questions for future research.

Future research should evaluate the practical implementation of the SEVBUS framework using objective operational data from different electric bus operators and geographical contexts. Such studies should assess its effectiveness as a diagnostic or planning tool and its applicability across operating environments. Longitudinal designs or explicitly specified structural models with independent outcome variables are also needed to test directional relationships and potential causal mechanisms among the SEVBUS dimensions.

5 Conclusion

This research developed and validated the Sustainable Electric Vehicle BUS Model (SEVBUS) as a multidimensional higher-order measurement framework for unscheduled electric bus operations. The model integrates four dimensions: Electric Bus Capability and Performance (EBCP), Service Scheduling and Utilization Planning (SSUP), Sustainability of Utilization and Operation (SUO), and Operational Cost Efficiency (OCE). Indicators and factors were developed through literature review and Delphi-based content validation, followed by EFA, CFA, and higher-order SEM measurement validation. The SEM analysis evaluated model fit and pairwise correlations among the four latent dimensions; no directional paths among EBCP, SSUP, OCE, and SUO were estimated.

The empirical analysis supports the reliability, convergent validity, discriminant validity, and acceptable fit of the proposed measurement framework. All six pairwise correlations among the four dimensions were positive and statistically significant (r = 0.38–0.61, all p < 0.001), indicating related but empirically distinguishable operational domains. These findings should not be interpreted as evidence that improvements in one dimension cause economic, environmental, or service outcomes in another dimension, or that service planning and resource utilization act as mechanisms that transform technological capability into such outcomes. Operational impact was not tested in the present study.

The principal contribution of this study is methodological: SEVBUS provides a validated construct architecture that integrates technological, service-planning, cost, and sustainability-related domains for unscheduled electric bus operations. Its theoretical contribution is limited to organizing these constructs within a coherent framework and documenting their empirical associations in the studied sample. Practical effectiveness has not yet been demonstrated. The framework may provide a basis for future diagnostic assessment and planning, subject to field implementation, external validation, and evaluation against objective operational outcomes.

5.1 Limitations and future research

One important limitation of this study concerns the technological context in the model. While the model reflects current technology of electric buses, it may not fully cover the rapid advancements in electric vehicle technologies. For example, high-density battery systems and ultra-fast charging technologies may reduce existing constraints related to driving range and charging duration, thereby potentially changing importance of factors within the Electric Bus Capability and Performance (EBCP) and Operational Cost Efficiency (OCE) dimensions. Similarly, advances in AI-enabled fleet management and predictive analytics may improve service scheduling, resource allocation, and operational coordination, particularly in dynamic and unscheduled service environments. Therefore, the relationships proposed in the SEVBUS model should be continuously revalidated and refined as electric bus technologies and operational practices evolve.

Another key limitation is that Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) were conducted sequentially on the same dataset of 400 respondents. Although the observed indicators were developed from the literature review and validated through the Delphi technique prior to quantitative testing, the use of a single dataset for both exploratory and confirmatory analyses may limit the robustness of factor-structure validation and increase the possibility of capitalizing on chance. In this research, the full dataset was retained for CFA and SEM due to the numbers of 48 observed indicators, 19 factors, and four latent dimensions, where reducing the sample size could have weakened the stability of parameter estimation. Nevertheless, this issue should be considered when interpreting the model-fit results and the generalizability of the proposed framework. Future studies should therefore employ split-sample validation, cross-validation, or independent validation samples to further confirm the stability, replicability, and robustness of the model.

A third limitation stems from the cross-sectional design and single period of data collection. Longitudinal validation would enable assessment of whether the associations among the four SEVBUS dimensions remain stable under changing operational and technological conditions. The present findings should therefore be interpreted as cross-sectional associations among latent constructs rather than as directional or causal effects. The use of questionnaire-based data may also introduce potential common method bias because the same respondents provided information on multiple constructs. Future studies should consider longitudinal designs, multi-source data, operational records, and explicitly specified predictor-outcome models when testing directionality.

Another limitation resides in the use of non-probability sampling and country-specific context. Because the sample was drawn from stakeholders involved in electric bus operations, the findings may not be fully generalizable to all electric bus systems in regulatory pillars or market structures. Accordingly, the findings should be interpreted within the operational, regulatory, and technological context of unscheduled electric bus services in Bangkok, Thailand. The applicability of the SEVBUS framework to other transport systems, service models, or geographical regions should therefore be confirmed through additional empirical validation. External factors such as electricity prices, charging infrastructure availability, renewable energy penetration, regulatory policies, subsidy schemes, and technological uncertainty were not explicitly incorporated into the present model. Future studies should examine how these external conditions influence the relationships among vehicle capability, service scheduling, operational cost efficiency, and sustainability performance.

To further gain insights on correlations on road conditions with the operation, future research could integrate Geographic Information Systems (GIS) to map spatial constraints affecting unscheduled electric bus operations, including road slope, traffic congestion, road surface conditions, charging station accessibility, and route-level infrastructure limitations. The analysis could support the localized or micro-scale sustainable mobility indices for evaluating operational efficiency. In addition, machine learning could also be employed to strengthen predictive robustness of the SEVBUS indicators. Algorithms such as Logistic Regression, Decision Tree classification, and Random Forest ensembles could be used to identify which operational predictors most strongly influence sustainability outcomes under dynamic and unscheduled service conditions.

Future research should continue to refine, validate, and implement the SEVBUS framework under evolving technological and operational conditions to evaluate its effectiveness as a practical assessment tool for sustainable electric bus management. Larger and more rigorous validation samples are needed to confirm the stability of the factor structure and inter-construct associations. Studies designed to establish directionality should use longitudinal data or outcome-based structural models, while GIS and computational techniques may be used to examine geographic and operational variation over time.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Francesca Cellina, University of Applied Sciences and Arts of Southern Switzerland, Switzerland

Reviewed by: Prapita Thanarak, Naresuan University, Thailand

Süleyman Nurullah Adahi Şahin, Sakarya University of Applied Sciences, Türkiye