Awareness without intention, attitude without action: a multi-method (SEM–ANN–fsQCA) study of Gen Z sustainable transport adoption in smart cities

1 Introduction

Sustainable urban development is a key focus in international policy frameworks, particularly in the SDG 11, which calls for the ‘creation of safe’, ‘resilient’, ‘inclusive and sustainable cities’ (Kutty et al., 2020). A key element of this goal is the development of sustainable urban transport solutions that can tackle environmental and traffic problems as well as increasing carbon emissions (Mavļutova et al., 2023). With rapidly increasing urbanization, particularly in developing countries, there is now a pressing need for efficient, sustainable, and technology-driven mobility systems, with low carbon emissions (Cao et al., 2024; Li et al., 2025). Sustainable transport, namely, public transport, ‘mobility-as-a-service’, and electric vehicles are considered crucial for achieving SDG and improving people’s quality of life (Anthony, 2023a,b).

In this dynamic urban context, a critical group that will determine our mobility future are Generation Z (Gen Z) individuals, whose changing behaviours are set to transform the transport sector (Chen M. et al., 2021; Chen X. et al., 2021; İlhan et al., 2022). Gen Z have developed a keen interest in technology and innovation due to having grown up in a digital era of global connection—attributes that are highly valuable in the development of smart cities (Garustowicz and Dawidowicz, 2025; Díaz et al., 2023). They can play a pivotal role for these initiatives through their perceptions, attitudes and intentions to adopt sustainable transport systems, and it is important to understand these factors (Acheampong and Siiba, 2019; Rad et al., 2023). Insight into the drivers of sustainable smart cities mobility for Gen Z is of interest not only for academic research but also for policymakers and urban planners to develop strategies and initiatives.

Although there has been increasing attention on adopting sustainable transport, linear analysis techniques, such as those used in ‘Structural Equation Modelling (SEM)’, has been overwhelmingly adopted to study behavioral relationships (Acheampong and Siiba, 2019; Nogueira et al., 2023; Rad et al., 2023). Although SEM can be used to test hypotheses derived from theories, and identify direct and indirect impacts, it has inherent limitations in modelling the complexity and non-linearity of human behavior (Guo et al., 2025). The use of linear methods limits the potential for extracting more profound insights into predictive relationships and could lead to simplifications of the underlying mechanisms same processes that lead to certain behavioral outcomes (Lyu, 2024; Niousha, 2025; Wu et al., 2024). Further, there is a lack of studies using advanced data mining techniques, such as ‘Artificial Neural Networks (ANN)’, which can capture ‘non-linear relationships’ and improve the accuracy of predictions (Rahman et al., 2021). Likewise, little research has explored the use of configurations approaches such as ‘fuzzy-set Qualitative Comparative Analysis (fsQCA)’ that can identify multiple causal paths with different combinations of variables that lead to a given outcome (Fiss, 2011; Woodside, 2013; Grunwald et al., 2025; Pham and Huynh, 2025).

To fill this gap, the current study takes a holistic multi-method view of Gen Z adoption of sustainable urban transport system in smart cities (Liu-Lastres et al., 2025). The research aims to understand the relationship among key psychological, social and environmental factors using SEM (Chen M. et al., 2021; Chen X. et al., 2021; Khalufi et al., 2025) while simultaneously taking advantage of the capabilities of the ANN to measure the predictive accuracy and importance of variables (Hair and Alamer, 2022; Hair et al., 2021; Palacios-Hernández and Lopez-Fernandez, 2025). Furthermore, it uses fsQCA to explore multiple configurations of conditions resulting in high levels of adoption (Grunwald et al., 2025). The study’s use of these interconnected analytical methods allows us to gain a greater depth of understanding of the phenomenon studied.

There has been growing academic interest in sustainable urban transport adoption; but previous research mostly focused on SEM in the analysis of behavioral relationships. SEM can only assess the linear causal relationships and does not have the explanatory power to show non-linear behavior patterns or multiple causal routes to the same behavior. The ‘adoption of sustainable transport modes’ is a complex phenomenon of behavior, which can be affected by various psychological, environmental, technical, and social conditions, and impacts that might in turn interact in non-linear and configurational manners. Therefore, uni-dimensional analytical method might not be a sufficient approach to understand the Gen Z approach to transport. With the intent to bridge such a gap, the present study combines SEM, ANN and fsQCA. SEM is used to determine the causal relationships, ANN to determine the non-linear effects and predictive importance, and fsQCA to determine multiple configurations that result in a high adoption intention. This study combines three theoretical angles: explanatory, predictive, and configurational to give a holistic understanding of the adoption of sustainable urban transport options among Gen Z in smart-city environments.

Based on this, the following research questions are considered in this study:

RQ1: What are the factors impacting Generation Z’s adoption intent and actual use of sustainable urban transport in the Indian smart cities?

RQ2: What are the factors that showed the greatest level of predictive power with regard to adoption intention when considering non-linear relationships?

RQ3: What are the combinations of psychological, environmental and social conditions that promote higher adoption intention for sustainable urban transport?

RQ4: Do explanatory, predictive and configurational approaches complement or contradict each other when explaining adoption of sustainable urban transport?

This study specifically contributes to ‘SDG 11’ which aims to “Ensure access to safe, affordable, accessible and sustainable transport systems for all, improving road safety, and promoting non-motorised transport.” The study assesses behaviours that drive an intention to use sustainable mobility from the perspective of Gen Z users, offering evidence to support policies that look toward enhancing the use of sustainable mobility within Smart City environments (Figure 1).

2 Literature review

2.1 Sustainable urban transport and SDG 11

Kutty et al. (2020) noted that sustainable transport in cities is critical to achieving SDG 11 since its implementation will ensure that cities and human settlements become more ‘inclusive’, ‘safe’, ‘resilient’, and ‘sustainable’. Urban development has exacerbated congestion, pollution and high carbon emissions, making it crucial that we adopt sustainable urban transport systems (Mavļutova et al., 2023). Sustainable urban transport includes ‘ecologically and economically efficient transport options’, such as ‘public transport’, ‘walking and cycling infrastructure’, ‘car-sharing schemes’, and ‘non-polluting vehicles’ (Gallo and Marinelli, 2020; Gulc and Budna, 2024). These transport systems help to achieve environmental sustainability, improve equity, accessibility, and quality of life (Rui and Othengrafen, 2023). Therefore, encouraging uptake of sustainable transport modes is crucial to achieving SDG 11 goals and prepare for resilient urban development (Bich and Dinh, 2024).

2.2 Gen Z and technology adoption behavior

Gen Z is a highly tech-savvy and socially minded generation that is set to have a critical influence on future transport trends (Rosenberg et al., 2025). Gen Z is a digital native who is highly proficient with smart technologies, apps and digital platforms that play a key role in urban mobility (Wawer et al., 2022). Not only convenience and efficiency but also environmental issues and social norms shape their travel behaviours. This makes Gen Z more open to the adoption of new and sustainable technologies than other age groups (Confetto et al., 2023; Theocharis and Tsekouropoulos, 2025). Knowing about the factors that drive their adoption behavior is therefore essential to design policy and technological tools to promote adoption of sustainable urban transport in smart cities (Anthony, 2023a,b; Mavļutova et al., 2023; Razmjoo et al., 2021).

2.3 Smart City mobility technologies: scope and rationale

Although this study situates Gen Z transport behavior within the broader smart city agenda, it is necessary to specify which categories of smart-city mobility technology are examined and why. The present research focuses on four interrelated categories of digitally enabled urban mobility infrastructure: (1) ‘mobility-as-a-service (MaaS) platforms that aggregate public transit, ride-hailing, and micro-mobility options within a single digital interface’; (2) ‘app-based ride-sharing and ride-hailing services’; (3) ‘digital and mobile ticketing or integrated payment systems for public transport’; and (4) ‘real-time transit information and vehicle-tracking applications’.

These four categories were selected for three reasons. First, they represent the principal digital touchpoints through which Gen Z users, characterized in the literature as digital natives (Wawer et al., 2022), interact with sustainable transport systems in daily practice, and therefore correspond most directly to the adoption behaviours this study seeks to explain, rather than to supply-side infrastructure that users do not directly experience. Second, they map onto the digital literacy construct operationalised in this study (items DL1–DL4; see Appendix), which asks respondents about their ability to use mobile applications, digital platforms, and online payment and ticketing systems for transport; the technologies examined are therefore embedded in the study’s own measurement structure rather than being an incidental backdrop. Third, this set of technologies is consistent with the priority mobility interventions promoted under India’s Smart Cities Mission and corresponds to the digital mobility services available in the smart-city contexts from which the sample was drawn, supporting the ecological validity of the “smart city” framing used throughout the manuscript.

The study does not examine supply-side smart infrastructure such as intelligent traffic signalling, sensor-based congestion management, or centralised transport control systems, since these operate at the level of city planning and are not directly experienced or evaluated by individual users; they therefore fall outside the scope of an individual-level adoption model of the kind developed here. This scope decision is stated explicitly so that readers can correctly calibrate the technological breadth of the smart city construct as it is used in this paper.

2.4 Theoretical foundations

2.4.1 Theory of planned behavior (ATT, PBC)

This research is underpinned by a range of theories that underpin sustainable urban travel (Davtyan, 2025; Kębłowski et al., 2022; Kent, 2021). This includes integrating behavioral science, technology research, and environmental science, to provide a holistic approach to informing Gen Z’s decision-making (Theocharis and Tsekouropoulos, 2025).

The ‘Theory of Planned Behavior (TPB)’ is a well-established model of individual intentions (Ajzen, 1991; Ajzen and Schmidt, 2020; Ali et al., 2023; Sun, 2019). TPB hypothesises that the key aspect in determining one’s behavior is attitude and ‘perceived behavioural control’ (Conner and Armitage, 1998; Ali et al., 2023; Shaaban and Maher, 2019). ‘Attitude is a person’s favorable or unfavorable evaluation’ of adopting sustainable transport, and ‘perceived behavioral control’ is the person’s ‘perceived ease’ or difficulty in adopting sustainable transport (Bamberg et al., 2003; Acheampong and Siiba, 2019; Ahn and Park, 2022; Tang et al., 2024). Understanding sustainable urban transport, a positive attitude and greater ‘perceived behaviour control’ are likely to increase the intention to adopt sustainable transport (Lanzini and Khan, 2017; Acheampong and Siiba, 2019; Nogueira et al., 2023). TPB has been used extensively in transportation and sustainability studies and is therefore a relevant framework to guide this research (Ali et al., 2023; Ciocîrlan and Rowe, 2025; Gajdzik et al., 2024).

The TPB is usually composed of ‘Attitude’, ‘Subjective Norm’ and ‘Perceived Behavioural Control’ but additions to the model are often made in the realm of sustainable mobility, introducing context-specific antecedents of the TPB that have environmental, technological, and behavioral aspects. The external antecedents of Social Influence are included in the TPB (normative), while Perceived Sustainability as well as Environmental Awareness and Innovation Resistance are added to the present study, forming behavioral evaluations and decisions regarding transport innovations in smart city settings. This is an expansion of the TPB, allowing more of a holistic analysis of sustainable urban transport adoption by Gen Z.

2.4.2 Innovation resistance/perceived sustainability

The general ‘Innovation Resistance Theory’ suggests that resistance to innovation decreases the intention to adopt and usage behavior (Ram and Sheth, 1989; Lyu et al., 2024; Chen et al., 2022). Sustainable urban transport (SUT) is, however, a different kind of technology adoption scenario since there are cases when users perceive environmental and societal benefits simultaneously, while at the same time feeling uncomfortable about other aspects like the convenience of the SUT service, the availability of the infrastructure, or the complexity of the technology (Talwar et al., 2023; Sarabia-Sanchez et al., 2021). That is why, innovation resistance in this study is considered as a multidimensional evaluative response and not as a negative barrier (Sarabia-Sanchez et al., 2021; Mavļutova et al., 2023). Based on the conflicting findings in the last set of sustainable mobility studies, the empirical relationship between innovation resistance and the uptake of sustainable transport is still open for discussion (Han, 2021; Vicente et al., 2020). The study therefore aims to explore these relationships in an empirical manner with an understanding that not all resistances may impede and that under certain contextual conditions, resistances may coexist with adoption behavior (Sarabia-Sanchez et al., 2021; Wang et al., 2020).

2.5 Hypotheses development

2.5.1 Perceived sustainability and behavioral outcomes

Perceived sustainability is a crucial factor in influencing people’s attitudes and behaviours toward sustainable practices (Bamberg and Möser, 2007; Hamidi and Zhao, 2020; Nogueira et al., 2023). Understanding the ecological and societal advantages of sustainable transport systems in urban areas will encourage the adoption of positive perceptions and behaviours of sustainability (Steg, 2005; Acheampong and Siiba, 2019; Nogueira et al., 2023; Topal et al., 2021). Moreover, ‘perceived sustainability’ can also affect adoption intentions and behavior of using sustainable transport (Steg and Vlek, 2009; Klöckner, 2013; Rad et al., 2023; Tang et al., 2024).

H1: Perceived sustainability positively impacts sustainability transport Attitude.

H2: Perceived sustainability positively impacts sustainable urban transport Self -Reported Actual Use.

H3: Perceived sustainability positively impacts Adoption Intention.

2.5.2 Innovation resistance and adoption behavior

Innovation resistance is the resistance of a user to adopt a technology, a product, or a new behavior due to factors like risks, uncertainty, compatibility with existing habits and complexity (Liao et al., 2025; Lyu et al., 2024). Most previous studies indicated that the negative effect of innovation resistance plays a role in attitude toward technology usage and technology adoption. Sustainable urban transport also presents a unique challenge in that individuals can at the same time be aware of the environmental and social advantages of sustainable transport without expressing concern about change, convenience, or technological complexity. Hence, the present study has an empirical look at the relationship between innovation resistance and sustainable transport adoption, without assuming that the relationship is always negative. Relationships are characterized by their direction and amount of relationship is determined empirically.

H4: Innovation resistance positively impacts perceived behavioral control.

H5: Innovation resistance positively impacts sustainable urban transport Self -Reported Actual Use.

H6: Innovation resistance positively impacts Adoption Intention.

2.5.3 Attitude toward sustainable transport

Positive attitudes toward sustainable transport play an important role both in making decisions to use sustainable transport and the use itself (Acheampong and Siiba, 2019; Si et al., 2019). According to the ‘Theory of Planned Behavior (TPB)’, people with positive attitudes are expected to use sustainable transport (Nogueira et al., 2023; Si et al., 2019).

H7: ‘Attitude toward sustainable transport’ positively impacts Self -Reported Actual Use.

H8: ‘Attitude toward sustainable transport’ positively impacts Adoption Intention.

2.5.4 Perceived behavioral control

‘Perceived behavioral control’ can be described as the ‘perception of the capability to engage in a certain behavior’ (Ali et al., 2023; Nogueira et al., 2023). Perceptions of control people feel they have and the resources they have to support this, can lead to sustainable transport use (Javaid et al., 2020; Nogueira et al., 2023; Si et al., 2019).

H9: Perceived behavioral control positively impacts Self -Reported Actual Use.

H10: Perceived behavioral control positively impacts Adoption Intention.

2.5.5 Adoption intention and Self -Reported Actual Use

Adoption intention has long been established as a predictor of behavior (Ahn and Park, 2022). People who have adoption intention for sustainable transport are more likely to adopt and use these ‘sustainable transport modes’ (Acheampong and Siiba, 2019; Rad et al., 2023; Si et al., 2019).

H11: Adoption intention positively impacts Self -Reported Actual Use of sustainable urban transport Actual Use.

2.5.6 Environmental awareness

‘Environmental awareness’ refers to people’s awareness of environmental problems and their commitment to engage in environmentally sustainable behaviours (Topal et al., 2021; Vázquez-Paja et al., 2024). The more awareness, the better the attitude, and the intention and ‘Self -Reported Actual Use’ (Acheampong and Siiba, 2019; Si et al., 2019; Tang et al., 2024).

H12: Environmental awareness positively impacts Attitude Towards Sustainable Transport.

H13: Environmental awareness positively impacts Self -Reported Actual Use.

H14: Environmental awareness positively impacts Adoption Intention.

2.5.7 Social influence

‘Social influence’ is the influence of people in friends’ networks, family and society (Arroyo et al., 2020). In youth, social influence may become a crucial factor in determining the ‘intention to use sustainable transportation’ as well as its Self -Reported Actual Use (Ahn and Park, 2022; Rad et al., 2023; Wang et al., 2020).

H15: Social influence positively impacts. Self -Reported Actual Use.

H16: Social influence positively impacts Adoption Intention.

2.5.8 Moderating role of digital literacy

Digital literacy is the capacity to access, comprehend, assess, and craft digital technologies and information effectively to benefit individuals (Ariyanti et al., 2025; Shin et al., 2021). In the world of the smart city, many sustainable mobility services like ride-sharing applications, digital ticketing systems, mobility-as-a-service (MaaS) and EV applications demand a degree of digital competence. Hence, there is a potential that more digitally literate individuals could be more inclined at realising positive attitudes toward sustainable transport into the intended behaviours and use. Thus, the authors suggest digital literacy as a ‘moderating variable’ which shall enhance the relationship between attitude and the outcomes of the adoption.

Digital literacy is the capacity to search, appraise and access digital technologies and information effectively. Smart-city mobility services are becoming more and more digital, with the use of various digital interfaces like mobile ticketing, ride-sharing apps, mobility-as-a-service platforms and real-time transport information platforms. Thus, there could be a positive relationship between digital literacy and translating favorable attitudes toward sustainable transport into behavior intentions and behavior use. Digital literacy is thus viewed as a facilitating factor or a moderator that can boost the impact of attitude to sustainable transport outcomes by making the use of technology for mobility more feasible (Table 1).

Table 1. Summary of hypotheses.

Source: Prepared by Author.

H17: ‘Digital literacy’ moderates the link between ‘Attitude and Self -Reported Actual Use’.

H18: ‘Digital literacy’ moderates the relationship between ‘Attitude and Adoption Intention’.

Gen Z is an especially significant audience in the Indian mobility landscape, as it is stepping into its most mobile demand window and has the most digital-fit, eco-consciousness among all the other generations. Preference for their transport modes, therefore, can influence the future demand for sustainable transport. Therefore, it is suitable to investigate factors that influence the adoption of transport among Gen Z to learn from it, which could not be fully applied in studies conducted on older generations.

3 Research methodology

3.1 Research design

The current research adopts a quantitative approach in order to evaluate the determinants of urban sustainable transport adoption by Gen-Z in smart cities. The study uses a survey-based approach for quantifiable data collection, allowing us to apply sophisticated statistical and mathematical analysis. To gain a holistic view of the problem, the study adopts a tri-method analysis approach, which includes SEM, ANN and fsQCA. SEM is used to analyze the proposed relationships in terms of validity using the latent variables and their direct and indirect influences.

ANN is used to boost the predictive power and map any non-linear effects that may exist but cannot be fully captured by SEM. Finally, fsQCA is used to discern the various combinations of causal factors that may result in high adoption behavior and thus providing a configurational view. This multimethod approach enables a comprehensive understanding by drawing inferences from explanatory, predictive and configurational analyses.

The theoretical approach into PLS-SEM was chosen, since the focus of this research is on the prediction of the models and the extension of the theories. This model features several latent constructs, moderation effects and a combined SEM-ANN approach. In addition, PLS-SEM is most applicable for exploratory and prediction-oriented research, whose purpose is mostly to ‘maximize the amount of explained variance of the endogenous constructs’.

3.2 Data collection

The survey was done in urban and semi-urban areas in India directly or indirectly linked to Smart Cities Mission activities with respondents belonging to Generation Z. The survey was carried out in the period from September 2025 till January 2026. Respondents had to be familiar with sustainable modes of urban transportation (metro rail systems, electric buses, app-based mobility systems, bicycle-sharing services, ride-share services, and electric vehicles and be familiar with their use and/or have previously used them). To remain in the sample, rural respondents needed to be involved in sustainable transport systems in urban centres that were close to them. This was an eligibility requirement to ensure that all the respondents were able to assess the ‘adoption of sustainable urban transport’ in the environment of a Smart City.

The study was targeted toward the respondents who were either living in cities part of the India’s Smart Cities Mission or travelling to these cities regularly. Mobility solutions evaluated included: metro rail systems, electric buses, app-based mobility platforms, bicycle sharing, ride sharing and EVs-serviced within a smart city setting. Respondents were asked to answer all the questions in relation to this transport system.

Non-probability convenience sampling approach was used. The questionnaire was administered via university channels, student unions, social media and via contacts. To screen the respondents and make sure that they were from among the Generation Z and have an affinity with sustainable urban transport systems. Questionnaires that had a considerable number of blank cells, some of which were not completed, and questionnaires where a similar response pattern was noticed were excluded, before being analysed. There were 433 useful responses for final analysis (Table 2).

Table 2. Demographic profile.

Source: Prepared by Author.

The demographic information suggests a balanced representation in terms of gender, with close to equal numbers of males and females. The highest proportion of respondents are aged 21–23, followed by 18–20, which suggests a strong presence of core Generation Z. A majority are university students, which is consistent with the age profile. When it comes to occupation, the number of students is the highest, followed by employed people, indicating a combined experience of student and early job experiences. Respondents are heavily urban and semi-urban, suggesting representation from regions with more convenient access to sustainable transport systems. Moreover, many respondents report they have used sustainable transport before, so it’s an informed sample when it comes to the study of ‘adoption behavior of sustainable transport’.

3.3 Measurement instrument

The measurement tool for this research was designed relying on scales that have been proven valid and reliable in previous studies. The questionnaire involved multiple constructs, such as perceived sustainability, innovation resistance, attitude toward sustainable transport, perceived behavioral control, environmental awareness, social influence, adoption intention, and usage, which were assessed by multiple items that were modified according to the context of ‘sustainable urban transport in smart cities’. Each item was rated on a ‘five-point Likert-scale where 1 = “strongly disagree” and 5 = “strongly agree”;’ this is a common scale used to measure perceptions and attitudes in behavioral research studies. The questionnaire was circulated among a few experts and academics to ensure that the items were relevant, understandable, and consistent with the context of the study. This facilitated response consistency and enabled the use of sophisticated analysis methods, including SEM, ANN, and fsQCA.

4 Data analysis and results

4.1 SEM analysis (PLS-SEM)

4.1.1 Measurement model evaluation

The ‘measurement model’ was used to test the ‘reliability and validity of the constructs’ of the model (Hair et al., 2019). ‘Cronbach’s alpha and composite reliability (CR)’ two well-accepted ‘measures of reliability’ were used to assess the ‘reliability of the constructs’. The findings show that all the constructs are ‘greater than the recommended value of 0.70’, indicating high reliability (Cheung et al., 2023). The ‘Cronbach’s alpha and CR values of all constructs’ were well above the acceptable threshold, supporting that the items used to measure the constructs are in line with their latent constructs (Cheung et al., 2023).

The ‘Average Variance Extracted (AVE)’ was used to evaluate the ‘Convergent Validity’, which refers to the ‘amount of variance that is captured by the construct in relation to its indicators’ (Hair et al., 2019). All the ‘AVE values’ were above the ‘minimum recommended value of 0.50’, suggesting that the constructs have satisfactory convergent validity (Cheung et al., 2023). This implies that the indicators of each construct are highly correlated, capturing the construct that they measure.

To test the ‘Discriminant Validity of the constructs’, both the ‘Heterotrait-Monotrait ratio (HTMT)’ and the ‘Fornell-Larcker criterion’ were applied (Radomir and Moisescu, 2019). The ‘HTMT ratios’ for all pairs of constructs were less than the ‘conservative cut-off of 0.85’ (Radomir and Moisescu, 2019), suggesting that the constructs are distinct from one another. The ‘Fornell-Larcker criterion’ was also met as the ‘Square root of AVE’ for all constructs exceeded their correlations with other constructs (Hilkenmeier et al., 2020). This indicates that our constructs display acceptable discriminant validity and are sufficiently separated in the model (Table 3).

Table 3. Construct reliability results.

Source: Calculated by Author.

This corroborates that the constructs have a ‘high degree of internal consistency as Cronbach’s alpha’, ‘Composite Reliability, and Dijkstra-Henseler’s rho values’ are above the ‘threshold value’. This shows that the items are a good representation of their corresponding constructs. Refer Appendix for Sources of Constructs considered from literature.

Several reliability coefficients were more than 0.95, which was considered very high internal consistency. Such values indicate good reliability, but past methodological research has suggested that very high reliability coefficients may also show item redundancy. For the present study, the measurement scales were some that have been familiar and extensively used, and were retained for content validity and comparison to previous studies. However, future research could be used to further reduce and/or reword some of the items for greater scale economy and sufficient representation of the construct (Table 4).

Table 4. Convergent validity (AVE).

Source: Calculated by Author.

The ‘AVE values’ for every construct are ‘higher than the threshold’, meaning that there is ‘strong convergent validity’. This means that ‘each construct explains a good part of the variance’ of the items (Table 5).

Table 5. Discriminant validity (HTMT and HTMT2).

Source: Calculated by Author.

The ‘HTMT and HTMT2 values’ are ‘lower than the recommended cut-off’, which suggests ‘good discriminant validity’. This suggests that all the constructs are empirically different from each other.

While all ‘HTMT were below the suggested threshold of 0.85’, higher values were found within the relationships of Environmental Awareness, Perceived Sustainability and Innovation Resistance. This pattern is theoretically possible since these constructs are conceptually linked in the sustainable mobility world. A person who cares a lot about the environment can also feel that on the one hand sustainable transport is good for the environment while on the other hand, he/she can feel concerned about the implementation of new mobility technologies. Although there were conceptual overlaps, all the ‘HTMT values’ were within the acceptable limit thereby endorsing the ‘discriminant validity’ (Table 6).

Table 6. Fornell-Larcker criterion.

Source: Calculated by Author.

The results from the ‘Fornell-Larcker criterion’ suggest that the ‘square root of the AVE’ of each construct is larger than the ‘correlations with other constructs’, reinforcing the ‘Discriminant Validity’ and indicating the constructs are distinct (Table 7).

Table 7. Factor loadings.

Source: Calculated by Author.

The indicator loadings are all greater than the recommended loading level, confirming indicator reliability. This supports item validity, ensuring each of the items used to ‘measure a construct represent its construct’.

4.1.2 Structural model evaluation

A ‘structural model analysis’ was carried out to establish how predictive the model was and to identify interactions between constructs. The analysis proves that the model had ‘adequate predictive ability’.

The results show perceived sustainability has a significant impact on attitude and intention to adopt and innovation resistance impacts perceived behavioral control. Both ‘perceived behavioral control’ and ‘social influence’ were found to positively influence adoption intention, which also shows a strong influence on Self -Reported Actual Use. Perceived sustainability has a significant impact on ‘attitude toward sustainable transport’, intention to adopt and use. Findings show that the less dangerous ‘sustainable transport’ is to the environment and the more socially desirable it is to the respondent, the more positive attitudes, intention to use and usage behavior they are likely to have. These findings support that although perceptions of sustainability have an attitudinal impact, they have a direct impact on behavior as well. Moreover, attitude has no significant effect on behavioral outcomes, which may mean that attitudes may not necessarily lead to behavior. Many hypotheses were supported, specifically those that relate perceived sustainability, ‘perceived behavioral control’, social influence and ‘adoption intention’, but some of the direct effects were not confirmed, suggesting that there may be an indirect relationship in place. Also, the moderating role of digital literacy was not significant, implying that it is not modulating the effect of attitude on behavior (Table 8).

Table 8. R² and model fit.

Source: Calculated by Author.

The R² values show the model has good to high levels of explanatory power, with adoption intention having the highest variance explain, followed by Self -Reported Actual Use, ‘attitude and perceived behavioral control’. This implies that the model has commendable predictive power for the endogenous key constructs. As for the model fit, the SRMR of the saturated model demonstrates that it fits well while the estimated model shows an acceptable fit, suggesting that the model fits well overall (Table 9).

Table 9. Model fit indices.

Source: Calculated by Author.

SRMR of estimated model: 0.1009, above the conservative value of 0.08. However, previous PLS-SEM studies indicate that ‘SRMR values’ below 0.10 may be acceptable in exploratory or prediction-oriented studies. Thus, the model fit should be taken as only marginal to accept and results should be judged along with the predictive performance indicators of the model.

The SRMR and R² and predictive performance indicator was used to assess the overall model. A good model fit was achieved (SRMR = 0.0209) for the saturated model. While the estimated model displayed an ‘SRMR value’ of 0.1009, which is slightly above the ‘conservative threshold value’ of 0.08 and hence should be interpreted with caution. Model fit was assessed not just based on the overall fit of the model but also in terms of the ‘explanatory and predictive capabilities of the model’, following the suggestions of prediction-oriented PLS-SEM approach. The model explained a great deal of the variance in ‘Adoption Intention’ (R² = 0.668), and moderate amounts of variance in ‘Self-Reported Use’ (R² = 0.582), Attitude (R² = 0.459) and ‘Perceived Behavioural Control’ (R² = 0.412). In combination, the findings indicate that despite having different suggestions from the global fit of the estimated model, the suggested model does have a satisfactory level of ‘explanatory and predictive power’ for the most important endogenous constructs (Table 10; Figure 2).

Table 10. Structural model results and hypothesis testing.

Source: Calculated by Author.

There is broad support for the hypotheses, with perceived sustainability, innovation resistance, perceived behavioral control and social influence effects prevailing in the structural model results. Usage is found to be significantly predicted by adoption intention, indicating that it is highly important in shaping behavior. But both attitude and environmental awareness are not directly related to either intention or Self -Reported Actual Use. And, the moderating impact of digital literacy is not observed.

4.2 ANN analysis

4.2.1 ANN model specification

To provide a more accurate model prediction and supplement the findings of SEM, an ANN analysis was used to check potential non-linear association and the importance of the predictors of sustainable transport adoption. The model featured perceived sustainability, innovation resistance, attitude toward sustainable transport, perceived behavioral control, environmental awareness and social influence (perceived to be important by the person) as predictor variables, chosen for their theoretical importance and empirical support, with adoption intention and Self -Reported Actual Use as dependent variables, respectively. ‘The feedforward multilayer perceptron model featuring an input layer, a hidden layer and an output layer were adopted for the ANN architecture and the number of neurons in the hidden layer was determined by trial and error to achieve a balance between accuracy and prevention of overfitting’. We used the backpropagation algorithm with a ‘non-linear function in the hidden layer’ and a ‘linear function in the output layer’ to train the model to reduce prediction errors. To ‘avoid overfitting and ensure the model would generalize well’, data were ‘split into training and testing sets’ and the model was validated on the testing data set.

The use of adoption intention as the only output variable for the ANN model considered in this study was justified since “adoption intention” was revealed as the key endogenous variable in the SEM model and the most theoretically significant targeted prediction variable for subsequent ANN analysis.

4.2.2 Model performance

The results of the ANN model is assessed based on performance metrics for predictive modelling such as the ‘Root Mean Square Error (RMSE) and R squared (R²)’. The results for determination of adoption intention show an RMSE value of 0.5837 for ANN, which means that there is a relatively small prediction error, and the model can predict the dependent variable well. The explanatory power of the ‘ANN model’ was shown to be high with an R² of 0.6294, suggesting that the input variables explain a significant proportion of the variance of adoption intention.

These results suggest that the ANN model offers good prediction accuracy and effectively deals with non-linear relationships between the variables, beyond what is captured by the SEM. Compared against the SEM results (adoption intention had an R² of 0.668), the ANN model demonstrates slightly weaker, but nonetheless quite similar, explanatory power for adoption intention, thus strengthening the multi-method approach (Tables 11, 12).

Table 11. ANN model performance results.

Source: Calculated by Author.

Table 12. ANN architecture.

Source: Prepared by Author.

The ANN model performance results show that the model performs well, with a low ‘root mean square error (RMSE) and a high R² value’. This indicates that the model exhibits good predictive ability for the variable “adoption intention,” and it explains a significant part of variance of the dependent variable. ANN complements SEM by identifying hidden non-linear relationships and predictor importance rather than necessarily improving explanatory power.

4.2.3 Sensitivity analysis

Sensitivity analysis was done to establish the relative weight of predictors used in the ANN models. Sensitivity analysis identifies the contribution of predictors to predictive ability indicated by SEM analysis. The findings reveal that environmental awareness is the most important predictor to predict adoption intention to use sustainable urban transport, as indicated by the highest relative importance using connection-weight and permutation importance approaches. This implies that peoples’ understanding of environmental issues is a key factor in predicting adoption of sustainable urban transport.

The second predictor, perceived sustainability, underlines the importance of peoples’ perceptions of the environmental and social benefits of sustainable transports. Innovation resistance and attitude toward sustainable transport also show average levels of importance, suggesting both positive and negative behavioral factors play a role in the adoption. Social influence is also demonstrated as significant, representing the effects of social pressure and perceptions of others in the context of Generation Z’s behavioral intentions.

On the other hand, ‘perceived behavioral control’ exhibits a relatively lesser level of importance in the ANN model than other predictors, which implies that while this predictor is significantly related to adoption intention in SEM, its contribution to the ANN prediction model is lesser (Table 13).

Table 13. ANN variable importance ranking.

Source: Calculated by Author.

As indicated by ANN analysis, environmental awareness is the key predictor of intention to act while sustainability is identified as the second strongest predictor. Innovation resistance, attitudes and social influence are moderately important predictor variables while perceived behavioral control has the least influence. The results indicate that environmental influences and perceptions are more important in predicting sustainable transport adoption than factors related to perceived control.

4.3 fsQCA analysis

4.3.1 Calibration of variables

In fsQCA calibration is an essential stage that converts data into a ‘fuzzy-set format’, with membership scores ‘between 0 and 1’. For this research, the calibration was done by direct method, which involves specifying three qualitative thresholds for each construct: ‘full membership, crossover point, and full non-membership’. These anchors determine the extent to which a case is considered a member of a set and play a key role in discerning relevant configurational patterns.

Based on theoretical and empirically determined thresholds, and since the data were from a ‘five-point Likert scale’, the membership thresholds were set as follows. The full membership threshold was set at a high value, reflecting a high level of agreement. The cross-over point, which refers to the point where there is uncertainty and cases are neither totally in nor out of the set, was set at the middle of the scale. The other threshold represented full non-membership, which was set at a low level, showing that the case disagrees with the construct. This ensures a fair and theoretically sound fuzzification of the data into sets.

Thresholds have been used to set up all variables, including perceived sustainability, resistance to change, attitude toward sustainable transport mode, ‘perceived behavior control’, ‘environmental awareness’, ‘social influence’, and ‘intention to use’. Through the calibration, necessary and sufficient conditions and multiple causal configurations for levels of adoption intention and Self -Reported Actual Use can be identified.

It should be noted that the original ‘five-point Likert scale’ was converted to ‘fuzzy-set membership scores’ using the ‘direct calibration method’ in the process of the fsQCA analysis. Consistent with the established guidelines for using fsQCA, all antecedent conditions were ‘calibrated from full membership to the crossover point indicating maximum ambiguity to full non-membership’, ranging from 4.5 to 1.5 and for the outcome variable—perceived sustainability, innovation resistance, attitude, perceived behavioral control, environmental awareness, social influence, digital literacy and adoption intention. Before we constructed the truth-table and derived the solution, we set the ‘frequency threshold’, the ‘consistency threshold’, and the ‘PRI consistency threshold as 3, 0.80 and 0.70 respectively’ to ensure the robustness and reliability of our configurational analysis. In addition, the intermediate theory was favored for interpretation as it serves to balance theoretical plausibility and empirical complexity without the drawbacks of the extremely simple parsimonious solution and extremely elaborate solutions which would otherwise be used. The methodological transparency and the possible replications of these calibration and analytical decisions also help pinpointing various causal configurations allowing sustainable urban transport adoption when considering the responses of Gen Z.

4.3.2 Necessity analysis

Concerning necessity, from the table, no condition appears necessary for the intention to adopt the use. But high consistency levels reflect the perceived importance of the issues related to ‘environmental awareness’, ‘perceived sustainability’ and ‘resistance to innovation’. The rest, such as ‘social influence’, ‘attitude’ and ‘perceived behavioral control’, are moderately important (Table 14).

Table 14. Necessary conditions results.

Source: Calculated by Author.

In the results, the consistency level of all the factors is less than the threshold needed to be considered as a necessary condition for adoption intention. The results indicate that while environmental awareness, perceived sustainability and innovation resistance have higher consistency values, they are not sufficient on their own. The other conditions, social influence, attitude and perceived behavioral control, are of moderate to low consistency (Table 15).

Table 15. FSQCA analytical parameters.

Source: Prepared by Author.

A ‘minimum frequency threshold of three’ and a ‘consistency threshold of 0.80’ was used to create the ‘truth table’. A ‘PRI consistency threshold of 0.70’ was also used to minimize contradictory configurations following recommended fsQCA procedures. Intermediate solutions were chosen for interpretation because they are intermediate in between a purely empirical and a purely theoretical approach and they are widely used in current fsQCA studies.

4.3.3 Sufficiency analysis

A sufficiency analysis was performed to determine ‘combinations of conditions sufficient to generate high intention to adopt and use sustainable urban transport’. This type of analysis looks at the interaction of multiple conditions to produce outcomes, rather than individual conditions as in necessity analysis. The findings suggest the presence of equifinality, which shows that there are many ways for adoption intention to be high. Perceived sustainability and perceived behavioral control are always key conditions, and social influence and environmental awareness are often complementary conditions that facilitate adoption behavior. On the other hand, the presence and absence of innovation resistance and attitude, toward sustainable transport in various configurations, points to a specific influence of these conditions, depending on their interaction with other conditions. As for the Self -Reported Actual Use, adoption intention appears in a majority of configurations, confirming the power of adopting intention in turning into actual behavior (Table 16).

Table 16. Truth table results.

Source: Calculated by Author.

Equifinality is clear from the truth table results with multiple configurations resulting in high adoption intention. The majority configurations consistently contain ‘perceived sustainability’, ‘perceived behavioral control’ and ‘environmental awareness’, suggesting their key contribution to the outcome. The configurations include social influence and innovation resistance in some but not all solutions, indicating that their effects are contingent. The high consistency scores across configurations suggest identified pathways have the ability to robustly explain adoption intention, with coverage scores varying across configurations suggesting that different combinations of conditions play varying roles influencing adoption intention.

4.3.4 Configurational solutions

The fsQCA-derived configurational solutions reveal more about the pathways to high adoption intention of sustainable urban transport—a principle of equifinality, where there are many pathways to the same outcome. These results show a strong focus on sustainability perception and perceived behavioral control, showing that people are likely to embrace sustainable transport if they view it as sustainable and have the power to adopt it.

Multilevel mindsets also play important role in many configurations, underlining the role of ‘perceived sustainability in promoting pro-environmental intention’ of the Generation Z. On the other hand, social influence plays a minor role, enhancing some configurations but not required in others. Likewise, resistance to innovation and attitude toward sustainable transport have conditional roles, with both appearing and not appearing across solutions, which may indicate that the effects of these causal factors depend on the presence or absence of other factors. The findings also show the presence of multiple paths, such as configurations where high level of perceived sustainability, environmental concern and ‘perceived behavioural control’ predict the intention to adopt regardless of being negative attitudes (Tables 17, 18).

Table 17. fsQCA solutions (configurations).

Solution 1: All conditions present, Solution 2: All present except social influence.

Solution 3: Innovation resistance absent; others present, Solution 4: Attitude absent; others present.

Source: Calculated by Author.

Table 18. Comparison of results across SEM, ANN, and fsQCA.

Source: Prepared by Author.

As it can be seen, there are convergences and divergences between analytical approaches: SEM, ANN, and fsQCA. One dimension which was perceived as important across all three methods is perceived Sustainability, showing there is consistency. But there was significant methodological difference in EA. In the SEM framework, it was not significant for predicting adoption intention or URLS, but it was the single strongest predictor in the ANN situation and a core condition in all the fsQCA solutions. The differences imply non-linear indirect and/or configurational effects that may possibly explain how Environmental Awareness impacts sustainable transport adoption if not addressed by linear modelling. Based on these findings, the nature of the complementary use of the three approaches – SEM, ANN, and fsQCA – is emphasized.

5 Discussion

This research gives a holistic approach to the issue of sustainable urban transport adoption by members of Gen Z using SEM, ANN, and fsQCA. Across the adopted methods, intention to adopt is the most important predictor of use, with a ‘high SEM path coefficient’ (β = 0.7628), suggesting a large effect of intention on use. The SEM model also has high explanatory power: adoption intention has a large explanatory power for the model (R² = 0.668); Self -Reported Actual Use has a moderate explanatory power (R² = 0.5819). The data confirm the major assumption of TPB, according to which intentions act as predictors of behavior in the field of sustainable mobility.

One of the most significant factors in the adoption of sustainable urban transport using all three analysis methods was perceived sustainability. SEM showed that perceived sustainability is very influential on intention to adopt (β = 0.5081) and sustainable transport awareness (β = 0.3817), meaning that it plays a very important role in the formation of behaviours. The second most significant predictor, with the second highest normalized importance value of 0.1905 out of 1.168 was perceived sustainability, supporting the result of the research. In addition, the fsQCA results showed that none of the low-consistency configurations that did not lead to a high value on the outcome were present, with all configurations that were present having a high value on the outcome having consistency values ranging from 0.9457 to 0.9591. This intersection of results from SEM, ANN and fsQCA gives strong evidence that attitude toward sustainability plays an important role as a predictor of sustainable transport adoption among the respondent Gen Z. These results concur with earlier studies which indicate that the perception of environmental, social and long-term sustainable outcome can substantially improve positive attitudes, ‘intentions to use and Self -Reported Actual Use’ of a sustainable transportation system.

Perceived control over behavior is also important, with significant impacts in SEM (β = 0.1385 on intentions to adopt) and appearing as a core condition in all fsQCA solutions. However, a relatively lower importance in ANN (importance = 0.1093) indicates that though it is significant in a linear statistical framework, its prediction power in a non-linear framework is relatively weaker. This suggests that perceived capability might not be a sole determinant of behavior, but needs to interact with other factors to achieve behavioral outcomes.

Another important finding of our study is that attitude is not significant. While a key component of the TPB, attitude is not found to be significant in its influence on adoption intention (β = 0.0469, p > 0.05) or use (β = 0.0357, p > 0.05). This is corroborated by the ANN findings of relatively low weight (0.1584/0.0325) of attitude, and fsQCA results, where it is only present in some configurations but not others. This finding implies that positive attitudes toward sustainable transport may not lead to adoption intentions or use, and points to a disconnection between attitudes and behavior.

There are mixed results for environmental awareness. While it significantly influences attitude in SEM (β = 0.3262), it does not directly affect adoption intention (β = 0.0153, p > 0.05) or Self -Reported Actual Use (β = 0.0117, p > 0.05). But it is the most important predictor (importance = 0.2259/0.2649) identified by ANN and a core condition in all configurations found in fsQCA. This contradicts the SEM results and indicates that environmental awareness might have indirect or non-linear impacts which could not be fully captured through SEM, hence the need for multiple analytical methods.

Innovation resistance and social influence have contextual moderately sized effects. SEM shows that innovation resistance significantly affects ‘perceived behavioral control’ (β = 0.6417) and adoption intention (β = 0.1869) while ANN shows it as moderately important (0.1585/0.0609). Likewise, social influence significantly impacts adoption intention (β = 0.1802) and Self -Reported Actual Use (β = 0.1374) in SEM and is moderately important in ANN (0.1574/0.0694). With fsQCA, both are included in the various configurations, suggesting that their importance varies depending on other conditions.

This positive relationship between ‘innovation resistance and adoption intention’ found in this study should be considered with caution. Resistance does not necessarily lead to adoption (in this case). This does not necessarily reflect, however, that the former is a reason to avoid while the latter is a justification to use, but that in the meantime the respondent perceives the concerns related to sustainable transport innovations and is still planning to take them up due to perceived environmental, social, and/or practical advantages. The findings might as well be an expression of the ambivalence of scepticism and readiness for adoption than any actual facilitating effect by resisting. Further studies should elicit the interactions of this relationship whether it is caused by suppression effects, multidimensional resistance constructs or through characteristics of sustainable urban mobility’s context.

The effect that digital literacy played as a moderator was not significant both for the adoption intention and self-reported use. A potential explanation is that there was not much variance among the level of digital competence among Gen Z respondents. Digital literacy value may not be the discriminatory difference when it comes to sustainable transport uptake among Gen Z as they are typically regarded as digital natives. The other reasoning might be that digital effectiveness has less impact in this context compared to other factors such as sustainability perception, accessibility, affordability and social influence.

The ANN model outputs a high correlation coefficient (R² = 0.6294) and low RMSE (0.5837) resulting in good prediction capability. While this value is slightly lower than our SEM model (R² = 0.668) this confirms the validity of the model and takes into account non-linear relationships. Further, the fsQCA results show varying-high-consistency configurations (0.9281–0.9614) with varying coverage (0.2470–0.4493), reaffirming the concept of equifinality. This also underlines that no only one condition is required for adoption (consistency < 0.90).

The result from the multi-method analysis that is important here relates to environmental awareness. The results of the SEM show that there is no direct link between the environmental awareness and the variables adoption intention or Self -Reported Actual Use, whereas ANN identifies that environmental awareness is the strongest predictor and fsQCA finds that environmental awareness is critical, forming a core condition in all high adoption configurations. This implies that the environmental consciousness has indirect, non-linear and context-dependent effects on the adoption of sustainable mobility. While it is assumed that Generation Z will have strong environmental concerns this may not directly serve as a behavioral intention as practical factors such as convenience, affordability, accessibility and social expectations may affect this. Thus, environmental awareness is used as an enabling condition that plays a role when it is paired with positive sustainability perceptions, behavioral control and social influence. This discovery suggests that combined use of complementary analytical tools provides a means to gain insight into behavioral mechanisms which might not be revealed by traditional linear analyses.

6 Conclusion

This study examined the factors shaping Generation Z’s adoption of sustainable urban transport in Indian smart cities in relation to SDG 11, using an integrated multi-method approach that combined SEM, ANN, and fsQCA. Across all three techniques, adoption intention emerged as the key predictor of use, confirming its pivotal role in bridging the gap between ‘intention and behaviour’. Perceived sustainability was the most consistently significant determinant, underscoring the importance of environmental and social benefits in shaping transport preferences, while ‘perceived behavioral control’ and ‘social influence’ also played meaningful roles in facilitating adoption. By contrast, attitude and environmental awareness were not significant predictors of adoption intention or use in the SEM model, despite emerging as important predictors in the ‘ANN model’ and as ‘core conditions in the fsQCA configurations’ – a divergence that points to interdependencies and non-linear, context-specific influences on Generation Z’s behavior that linear SEM alone does not capture. The absence of a moderating role for digital literacy further suggests that technological readiness is not a critical constraint for this cohort, while the configurational analysis confirms equifinality, showing that multiple alternative combinations of psychological, environmental and social conditions can lead to high adoption intention.

Taken together, these findings show that promoting sustainable urban mobility among Generation Z requires a holistic approach that looks beyond attitude change to encompass sustainability perceptions, behavioral control, and social context. By integrating explanatory, predictive and configurational perspectives, this study contributes both theoretical insight and an empirical evidence base for the design of interventions and policies that can better promote sustainable transport use among Generation Z in smart cities. The study’s limitations, directions for future research, and detailed theoretical and practical implications are set out below.

7 Limitations of the study

Several limitations should be noted in interpreting the results of this study. First, the cross-sectional design does not allow inferences of causality or temporal relationships among the constructs. Second, all variables were measured via self-reported questionnaire responses, and although steps were taken to address common method bias, it cannot be fully ruled out. Third, non-probability convenience sampling was used, which may limit the representativeness of the sample and the generalisability of findings to the wider Generation Z population. Fourth, actual use was captured through self-reported measures rather than objectively observed behavior. Fifth, the analysis did not account for city-level contextual factors that could affect sustainable transport uptake, including transport infrastructure attributes, service quality, reliability and affordability, urban density, or local policy settings. Sixth, the sample was drawn from Generation Z respondents in selected Indian smart-city contexts, so the results may not extend to other demographic groups or geographic settings. Finally, this study did not examine the political economy of transport policy implementation – including institutional capacity, stakeholder contestation, and governance variation across Indian Smart Cities – which constrains the extent to which the findings can speak to implementation-level dynamics.

8 Future research

Building on these limitations, future research should adopt multi-city comparative designs, objective behavioral measures, and probability-based sampling methods, ideally within a longitudinal framework, to strengthen the evidence base for the relationships identified here. Combining individual-level adoption modelling of the kind used in this study with policy-level analysis of institutional capacity, stakeholder contestation, and governance variation across Indian Smart Cities would offer a more complete picture of how political and administrative factors condition sustainable transport adoption. The ambivalent relationship observed between innovation resistance and adoption intention also warrants further investigation – future studies could examine whether this reflects suppression effects, multidimensional resistance constructs, or context-specific features of sustainable urban mobility. Finally, given the very high reliability coefficients observed for several constructs in this study, future work could revisit and refine the measurement items to reduce potential redundancy while preserving content validity.

9 Implications

9.1 Theoretical implications

Our research makes a series of contributions to the sustainable transport adoption theory, in the context of SDG 11. First, it builds on the TPB by showing that conventional assumptions about attitude may not hold true for sustainable transport behaviours in Generation Z. Although many previous studies emphasise the role of attitude in predicting ‘behavioural intention’ (Ali et al., 2023; Si et al., 2019; Nogueira et al., 2023), this study shows that attitude does not play a significant role in predicting the intention to adopt sustainable transport or how it is used. This contradicts the traditional TPB model and implies that the success of sustainable transport initiatives may be more likely explicable by practical and situational considerations rather than by attitudes. Our results are in line with recent research showing that behaviours are increasingly driven by contextual and practical factors rather than being exclusively explained by cognitive factors (Hamidi and Zhao, 2020; Tang et al., 2024).

Second, this research adds to the field of sustainability and transport studies by finding that perceived sustainability is an important factor influencing transport decisions. This is in line with previous research (Acheampong and Siiba, 2019; Rad et al., 2023; Vicente et al., 2020), which reported the important impact of perceived sustainability on behavioral intention and actual usage. But this research builds on this evidence by showing that the role of ‘perceived sustainability’ is not contingent upon the analytical methodology used (SEM, ANN and fsQCA). This supports the view that perceived sustainability is not just an “add-on” but a central element in the formation of ‘pro-environmental behaviours’, particularly in smart cities and sustainable mobility solutions (Anthony, 2023a,b; Mavļutova et al., 2023).

Third, this research advances ‘Innovation Resistance Theory’ by showing that innovation resistance can have varying effects depending on the context. While existing studies regard innovation resistance primarily as an impediment to adoption (Chen et al., 2022; Xue et al., 2024; Liao et al., 2025), this study shows that its role in affecting adoption is more complex and contingent on other variables. While resistance has a ‘negative influence on adoption’ in certain contexts, in others the influence is reduced by high levels of enabling factors, such as perceptions of sustainability and environmental concern. This suggests the importance of re-defining innovation resistance as part of a complex network of variables rather than a simple block of behavior.

Additionally, the study contributes to theory by showing the shortcomings of linear approaches in explaining behavior. The SEM findings confirm some relationships, but the ANN findings show non-linear relationships, in particular, the most effective predictor (environmental awareness), which demonstrates an insignificant direct effect in the SEM framework. Such evidence extends existing literature that supports the application of sophisticated analytical methods to discover certain non-obvious relationships between individuals’ behavior (Rahman and Al-Ameri, 2023; Guo et al., 2025; Lyu, 2024; Niousha, 2025). It also reveals that environmental awareness could have an indirect or non-linear influence that is not accurately modelled with linear approaches, hence necessitating a more holistic approach in research on human behavior.

Fifth, the use of fsQCA helps advance configurational theory by showing that the adoption of sustainable transport exhibits equifinality in that multiple sets of conditions are sufficient to achieve the same outcome. This observation is in line with other studies that support the use of configurational perspectives to analyze complex social processes (Chong, 2013; Grunwald et al., 2025; Pham and Huynh, 2025). The lack of a necessary condition supports the notion that not only is behavior not a direct result of individual conditions, but also the impact of each of these conditions is context dependent.

Fourthly, and lastly, this research makes an important contribution from the point of methodology as far as the integration of SEM, ANN, and fsQCA. Whereas previous research has been largely confined to individual explanatory and predictive approaches, this study highlights the benefits of integrating hybrid explanatory, predictive and configurational approaches to shed a wider light on behavior. This integrated analysis approach echoes calls (Guo et al., 2025; Palacios-Hernández and Lopez-Fernandez, 2025) in the literature to develop more rigorous and integrated analytical approaches and lays the groundwork for further research on sustainable transport and smart cities.

9.2 Practical implications

While the recommendations below translate the study’s findings into action points for policymakers, it is important to situate them within the political and institutional realities of urban transport governance. Sustainable transport interventions are rarely value-neutral technical exercises: they involve competing claims on scarce municipal budgets, negotiation with entrenched private and para-transit operators (for example, auto-rickshaw unions and private bus operators) whose livelihoods may be affected by policy shifts, and uneven administrative capacity across Indian smart cities to plan, fund, and sustain digital mobility infrastructure. The practical implications that follow should therefore be read as directions requiring political negotiation, institutional coordination, and phased implementation, rather than as straightforward technical prescriptions.

This study has several practical implications for policy-oriented analysts, urban planners and transportation service providers who are interested in facilitating sustainable urban transport as per SDG 11. First, the high impact of perceived sustainability on intention to adopt and use of sustainable transportation indicates that policymakers should promote the ‘environmental and social: benefits of sustainable transportation’. Government campaigns and policies should promote the ‘environmental and social benefits of sustainable transport’, such as ‘reducing pollution and improving urban living quality, as well as long-term sustainability’ (Anthony, 2023a,b; Bich and Dinh, 2024). Emphasizing these advantages can improve perceptions and boost uptake by Generation Z.

Second, the crucial impact of environmental concern, especially as the most important predictor in the ANN model, suggests specific efforts in education and awareness raising. Both governments and educational bodies should embed sustainability education in their programs and campaigns to encourage environmentally friendly behavior through online portals. As Gen Z is very sensitive to the environment (Chen M. et al., 2021; Chen X. et al., 2021; Confetto et al., 2023), raising awareness can be translated into increased sustainable transport uptake.

Third, ‘perceived behavioural control’ is key, so enhancing the accessibility, affordability and ease of use of sustainable transport options are critical. Governments should focus on infrastructure development, such as developing public transport routes and providing easy access and access to digital platforms. Improving convenience and practicality can increase people’s perceived behavior control over using sustainable transport (Javaid et al., 2020; Nogueira et al., 2023).

Fourth, our findings on social influence emphasise the importance of social norms and peer influence. Public transport providers and marketers should need to use social media, promoting through respected influencers and engaging in community-based programs to ‘encourage the use of public transport’. The creation of positive social norms and endorsement from peers can increase uptake among young people (Arroyo et al., 2020).

Fifth, attitude has a marginal impact on transport usage, which indicates that fostering positive attitudes may not be effective in promoting transport usage. To encourage adoption, more effective strategies are likely to be pragmatic approaches, such as enhancing the ‘quality, reliability, and convenience of the services’. This suggests that policy-makers should focus less on attitude promotion and more on improvements in transport services to promote transport use.

Sixth, the configurational results from fsQCA suggest there’s no universal approach to encourage sustainable transport use. Various user segments might be sensitive to various combinations of factors, including perceptions of transport sustainability, environmental concern, and social factors. As a result, policy makers should take an adaptive approach that considers designing strategies that target different needs of Generation Z users (Grunwald et al., 2025; Pham and Huynh, 2025).

Lastly, the lack of moderating effect of digital literacy indicates that the level of Generation Z’s digital competence may not be a critical factor. But this does not mean digital platforms are not important in smart cities. Public transport operators should further explore various digital services, such as ‘mobile apps’, ‘real-time tracking’ and ‘integrated payments’ that can enhance convenience and convenience (Ahn and Park, 2022; Teng et al., 2024).

Lastly, because of its digital-first, there’s an equity risk linked to a few of the above recommendations: mobile ticketing, real-time information apps, ride-hailing integration. The results of this study do not indicate significant involvement of Digital Literacy as a moderator among the Gen Z sample, but this should not be considered as a generalisable outcome, due to the existence of potential Digital Exclusion among other socioeconomic and age groups. Policymakers aiming to deliver smart mobility solutions should introduce digital services in combination with non-digital service to prevent an aggravation of the transport inequality. More broadly, technology adoption in this sense is not just a roll out exercise, but a politically negotiated process, which is influenced by priorities given to investments in infrastructure, capacity of urban governments and a differential bargaining position for providers of formal and informal transport.

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.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frsc.2026.1867418/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Angelo Genovese, University of Milan, Italy

Reviewed by: Mohammad Azmoodeh, Babol Noshirvani University of Technology, Iran

Fatma Kürüm Varolgüneş, Bingöl University, Türkiye

Danupon Sangnak, Kasetsart University, Thailand

Divya Singh, National Institute of Technology, Durgapur, India