Investigating the preferences for autonomous vehicle apply in European road transport: a binary logit model

Building upon the findings of the above-reviewed literature and research on driving automation, an online survey was developed and conducted to gather questionnaire responses. The objective was to analyze and derive insights into factors influencing the adoption of fully automated vehicles. This includes the socio-demographic characteristics of the sample, as well as a detailed description of the data processing methods applyd to create the database, which served as the foundation for the analysis and the formulation of the conclusions discussed in this study.

First, we present the descriptive statistics of the sample collected. The questionnaire responses collected from various European countries are analyzed to then present the derived modeling results based on the online survey. The primary objective is to highlight the characteristics of the respondents, whose input forms the basis of the results. These individual characteristics encompass the physical, social, and behavioral attributes of the sample, providing critical insights into the adoption of AVs. The survey findings are based on a sample of ${\boldsymbol{n}}$=235 responses from twelve countries across Europe.

The statistical analysis of the questionnaire data is summarized as follows, with results classified into four categories according to the questions posed to the participants:

Socio-Demographic Characteristics: Variations in the sample are analyzed based on gfinisher, income, age, and employment status (Table 1).
Mobility Behavior: Insights into the adequacy, satisfaction, transport mode, and trip purpose of transportation among respondents (Table 2).
Perception of Autonomous Vehicles: Results cover the perception of applying autonomous vehicles in urban and highway settings, attitudes toward autonomous public transportation, opinions on the cost of autonomous vehicles, and concerns about personal data management (Table 3).
Knowledge about Autonomous Vehicles: Statistics reflect respondents’ familiarity with autonomous vehicle technology, its usage, and its potential influence on their decisions (Table 4).

Table 1 Sociodemographic information.

Table 2 Mobility behavior information of the respondents.

Table 3 Respondents’ Perception of Autonomous Vehicles.

Table 4 Respondents’ knowledge about Autonomous Vehicles.

For most questions, respondents evaluated scenarios applying a Likert scale ranging from 1 for the most negative answers (“totally dissatisfied”, “completely inadequate”, “absolutely not”, “complete ignorance”, “totally disagree”) to 5 for the most positive answers (“totally satisfied”, “completely adequate”, “absolutely yes”, “excellent knowledge”, “totally agree”).

First, the socio-demographic characteristics of the sample are summarized in Table 1. The gfinisher distribution was balanced, with no significant differences between male and female respondents. The age distribution of the sample reflected a diverse representation of the typical workforce, with individuals over 65 years old comprising just 1% of the total respondents. Employment status revealed that over 65% of respondents were engaged in full-time work, while 30% identified as students, and approximately 2% were retired. Regarding economic status, 64% of respondents reported a monthly income below 1500€, whereas only 9% earned more than 3500€ per month.

Regarding the mobility behavior of the respondents, Table 2 reveals that approximately 60% commute for work or educational purposes. Additionally, 71% rely on either private vehicles or public transportation for their daily mobility necessarys, indicating a direct influence of automation on their choices. Dissatisfaction with available transportation options is observed among 22% of respondents, who rated their satisfaction as low (responses 1 to 2), while 40% expressed a neutral perspective (response 3). Concerning the adequacy of public transport in their local neighborhoods, dissatisfaction levels increase to 46% (responses 1 to 2).

Regarding respondents’ perception of AVs, Table 3 reveals that the majority of respondents hold a positive perspective on the apply of driverless public transport, with 88% responding “Yes” or “Maybe,” while only 12% expressed complete opposition. Participants indicated a higher level of trust in highway AV apply compared to city center apply. Specifically, 29% displayed a somewhat positive to positive attitude (responses 4 to 5) toward applying autonomous vehicles in city centers, while 41% expressed similar views for highway apply.

In terms of affordability, 71% of respondents confirmed that the cost of autonomous vehicles is a critical consideration (responses 4 to 5), and they should be accessible to all citizens. Finally, 38% of respondents indicated a high level of confidence (responses 4 to 5) that the protection of personal data significantly influences their decision-creating process.

Table 4 displays a general lack of familiarity with respondents’ knowledge of autonomous vehicle technology. Specifically, 44% reported having little to no knowledge of AV technology, while 23% indicated full or near-complete understanding. This limited awareness appears to influence decision-creating, as 29% of respondents (responses 4 to 5) stated that their lack of knowledge prevents them from considering the apply of an autonomous vehicle.

When it comes to exposure to automated vehicles, the sample was divided: 55% reported never having driven such a vehicle, while 45% had prior experience. Regarding perceptions of safety, 23% of respondents expressed confidence or near confidence that conventional vehicles are safer than autonomous ones, whereas 34% believed the opposite, favoring autonomous vehicles in terms of safety.

In conclusion, the statistical analysis of questionnaire responses revealed several indicative trfinishs relating to the broader relationship between socio-demographic characteristics and perceptions of autonomous vehicle technology. For example, respondents who reported higher levels of familiarity with AV technology were more likely to express a willingness, or at least openness, to apply driverless transportation options.

These findings, while suggestive of meaningful patterns, are derived from a modest sample of 235 individuals from diverse European countries. However, the sample reflects uneven representation across some demographic categories, including age and income levels, which limits the generalizability of specific subgroup insights. Therefore, the observed trfinishs should be interpreted as sample exploratory, not definitive. Further research with larger and more balanced samples is necessaryed to validate and expand upon these early observations.

Regarding the Preference for Autonomous Vehicles, we examine the willingness of respondents to apply autonomous vehicles, based on their evaluation of various factors. The questionnaire questioned respondents to compare a series of scenarios in which these factors occurred simultaneously and to indicate whether they would choose to apply an autonomous vehicle under those specific conditions. Five key factors were identified as most significant based on the literature review: protection of personal data, road infrastructure, legislation, vehicle cost, and passenger safety. For each of these factors, respondents were presented with scenarios in which the examined factor had a negative outcome while the others remained positive. The statistical results of the sample’s decisions, considering all possible combinations of these factors, are illustrated in the following figures.

As displayn in Fig. 3, the legislative framework consistently had a greater influence on decision-creating than the other factors, with the difference between yes and no answers averaging around 20% in three instances. The only exception to this pattern occurred in the case of safety, where the legislative framework’s impact was still significant but with a compacter difference of 6%.

**Fig. 3: Scenario assuming an insufficient legislative framework.**

Figure 4 demonstrates that road infrastructure is a critical factor influencing the choice to apply an autonomous vehicle, with a notably large difference. In most scenarios, this difference in choices exceeds 40%, except when evaluating safety, where it reduces to just over 20%.

**Fig. 4: Scenario assuming insufficient road infrastructure.**

Figure 5 reveals notable variability in the importance of the cost of AVs. Respondents prioritize safety and road infrastructure over cost, regard the legal framework as equally important, and consider privacy to be of lesser significance. Interestingly, while other factors are ranked higher in these scenarios, prior responses from Table 3 displayed that 71% of respondents were confident or nearly confident (responses 4 to 5) that affordability is essential.

**Fig. 5: Scenario assuming autonomous vehicles are financially unaffordable.**

Figure 6 highlights passenger safety as the most critical factor by the respondents, with overwhelming support across all scenarios presented.

**Fig. 6: Scenario assuming passenger safety is not guaranteed. In this scenario, it is assumed that passenger safety is not ensured.**

Additionally, Fig. 7 demonstrates that data privacy protection emerges as the preferred factor in most scenarios, though the differences are relatively minor, ranging from 5% to 10%. This suggests a divided attitude among the respondents on the importance of data privacy.

**Fig. 7: Scenario assuming data privacy protection is not ensured.**

In summary, the statistical analysis of scenarios involving the five examined factors underscores passenger safety as the primary criterion in decision-creating. Across all cases, the preference for safety exceeded 50%, with the difference reaching 66% when compared to data privacy protection.

Road infrastructure emerged as another critical factor, with an observed difference of 40% between “yes” and “no” responses. Similarly, the legislative framework consistently ranked higher than other factors in all scenarios, although the variations were less significant than those observed for passenger safety and road infrastructure.

In contrast, the sample displayed a divided opinion on vehicle cost and data privacy protection. Regarding AV cost, respondents were not willing to purchase an autonomous vehicle even if their data privacy was assured. However, they were more inclined to consider purchasing one despite high costs if passenger safety was guaranteed (22% difference) or if there was adequate road infrastructure (8% difference). These findings align with the prioritization of safety and infrastructure as key factors.

Lastly, opinions regarding data privacy protection were divided across all scenarios. In cases where the difference between positive and negative responses exceeded 10%, the legislative framework played a decisive role, with the majority prioritizing legislation over data privacy (12% difference). Similarly, road infrastructure was considered more important than data privacy, with a difference of 14%. These results highlight the nuanced preferences of respondents in evaluating autonomous vehicle adoption.

Building upon the statistical analysis of the responses, the next step involves Modelling applyr Preferences for Autonomous Vehicles. Using the data collected from $n=235$ online questionnaires, the primary objective is to analyze and evaluate the factors influencing European citizens’ decisions to adopt or apply autonomous vehicles. Below, we present the applied modeling approach and results.

Once participants completed the online questionnaire, the necessary data was gathered for modeling purposes. Using questionnaire responses, variables related to respondents’ personal information, travel behavior, and knowledge of autonomous vehicles were examined to develop a binary logit model. Each response in the survey represented an individual’s evaluation of various choice scenarios, with each alternative clearly defined.

The depfinishent variable ($Y$) was defined as the respondent’s preference for applying an autonomous vehicle or not for their daily commute. Specifically, Y=1 indicated a willingness/preference to adopt autonomous vehicles, while Y=0 signified a refusal. To assess the impact of the factors under consideration, the utility function was calculated. The data collected through the online survey was transformed into a long-format dataset, enabling the calculation of the utility function applying the R programming language.

A correlation analysis was subsequently performed to examine the relationships among the variables under consideration. This method evaluates the strength and direction of associations between variables, with correlation coefficients ranging from -1 to +1. Values near zero indicate no or very weak correlation. In this study, all variables demonstrated either zero or very weak correlations with each other (Fig. 8). A few expected relationships displayed relatively higher correlations, which serve to validate the internal consistency of the responses. These include:

The age group of 18–25 (age18_25) and student as professional status (ps_unistudent), with a correlation of 0.77.
The age group of greater than 65 (age_gr65) and retired as professional status (ps_retired), with a correlation of 0.70.
Travel purpose for education (Education) and student as professional status (ps_unistudent), with a correlation of 0.70.

**Fig. 8: Correlation matrix of variables under consideration.**

After evaluating multiple scenarios and combinations of factors associated with the examined variables and coefficient types, the utility function, as presented in Eq. (2), was determined to represent the model formula:

$$\launch{array}{l}Logit\left(p\right)={\text{ln}}\left(\frac{p}{1-p}\right)\\ ={\beta }_{0}+{\beta }_{AVknowledge3}* A{V}_{knowledge{3}}+{\beta }_{AVknowledge}* A{V}_{knowledge5}\\ +{\beta }_{drive{n}_{AV}}* drive{n}_{AV}+{\beta }_{AVsaferthanC{V}_{3}}* A{V}_{saferthanC{V}_{3}}+\,{\beta }_{AVsaferthanC{V}_{4}}\\ * A{V}_{saferthanC{V}_{4}}+{\beta }_{AVsaferthanC{V}_{5}}* A{V}_{saferthanC{V}_{5}}+{\beta }_{AVtrusthighwa{y}_{5}}\\ * A{V}_{trusthighwa{y}_{5}}+{\beta }_{dataprivac{y}_{AVpreference4}}* dataprivac{y}_{AVpreference4}\\ +{\beta }_{dataprivac{y}_{AVpreference5}}* dataprivac{y}_{AVpreference5}+{\beta }_{autonomou{s}_{P{T}_{Europe}}}\\ * autonomou{s}_{P{T}_{Europe}}+{\beta }_{Gfinisher}* Gfinisher+{\beta }_{age{36}_{45}}* age{36}_{45}+{\beta }_{incom{e}_{gr5000}}\\ * incom{e}_{gr5000}+{\beta }_{ag{e}_{gr65}}* ag{e}_{gr65}+{\beta }_{Mc}* Mc+{\beta }_{Entertainment}\\ * Entertainment+{\beta }_{Vehdr}* Vehdr+{\beta }_{Safety}* Safety+{\beta }_{Infrastructure}\\ * Infrastructure+{\beta }_{Framework}* Framework\finish{array}$$

(2)

Where: $p$: is the probability of applying an Autonomous Vehicle,

${\beta }_{0}$: alternative specific constant (or intercept) of the utility function,

${\beta }_{i}:$ coefficient of variable i (beta parameters),

${{\rm{AV}}}_{{\rm{knowledge}}3}$: 1 if the respondent has neutral stance on their knowledge (level 3) of AVs (dummy variable),

${{\rm{AV}}}_{{\rm{knowledge}}5}$: 1 if the respondent has excellent knowledge (level 5) of AVs (dummy variable),

${{\rm{driven}}}_{{AV}}$: 1 if the respondent or their relative has ever driven a car with automation elements (binary),

$A{V}_{{safe}{{rthanCV}}_{3}}$: 1 if the respondent has a neutral stance (level 3) that driving an AV is safer than a Conventional Vehicle (dummy variable),

$A{V}_{{safe}{{rthanCV}}_{4}}$: 1 if the respondent positively believes (level 4) that driving an AV is safer than a Conventional Vehicle (dummy variable),

$A{V}_{{safe}{{rthanCV}}_{5}}$: 1 if the respondent totally agrees (level 5) that driving an AV is safer than a Conventional Vehicle (dummy variable),

$A{V}_{{trus}{{thighway}}_{5}}$: 1 if the respondent totally trusts (level 5) an AV to drive on a highway (dummy variable),

${dataprivac}{y}_{{AVpreference}4}$: 1 if the respondent positively believes (level 4) that data privacy issues affect the preference for an AV (dummy variable),

${dataprivac}{y}_{{AVpreference}5}$: 1 if the respondent totally agrees (level 5) that data privacy issues affect the preference for an AV (dummy variable),

${autonomou}{s}_{P{T}_{{Europe}}}$: 1 if yes for knowing that Autonomous PT modes are already/ currently being applyd in Europe (binary),

${Gfinisher}$: 1 for female, -1 for male, and 0 for diverse,

${age}3{6}_{45}$: 1 if the respondent’s age is between 36-45 years old (dummy variable),

${incom}{e}_{{gr}5000}$: 1 if the respondent’s monthly income is greater than 5000€ (dummy variable),

${ag}{e}_{{gr}65}$: 1 if the respondent’s age is greater than 65 years old (dummy variable),

${Mc}$: 1 if the respondent applys a Motorcycle as their main transport mode in their everyday life (binary),

${Entertainment}$: 1 if the respondent’s main trip purpose of transportation in their everyday life is for entertainment (binary),

${Vehdr}$: 1 if the respondent is driving a vehicle as their main transport mode in their everyday life (binary),

${Safety}$: 1 if the respondent would prefer an AV when the passenger safety is guaranteed by the car indusattempt (binary),

${Infrastructure}$: 1 if the respondent would prefer an AV considering sufficient road infrastructure (binary), and

${Framework}$: 1 if the respondent would prefer an AV considering an adequate legislative framework (binary).

In Table 5, the coefficients of the parameters included in the final utility function of the binary logistic regression model are presented. The analysis of the utility function results was conducted by examining the coefficients, their signs, and their significance levels, as follows:

Coefficient: In the discrete choice model (here, a binary logistic regression model), the coefficient represents its impact on the overall utility. A negative coefficient indicates that an increase in the corresponding factor reduces utility, while a positive coefficient suggests that an increase in the factor increases utility. Comparisons between coefficients are possible when binary coding is applied.
Significance Level (P[ | Z | >z]): This statistical parameter assesses the results of the Wald test, determining whether a parameter’s effect equals to zero. It evaluates the relevance of each variable in the utility equation of the discrete choice model. The significance level ranges from 0 to 1, where lower values indicate greater certainty about the variable’s impact. For instance, a value of 0.001 indicates that the coefficient is statistically significant with 99% confidence and a 0.1% error margin.

Table 5 Binary logistic regression model – Utility function coefficients

The results in Table 5 demonstrate that all the characteristics included in the model for autonomous vehicle adoption are statistically significant at a 99% confidence level. Furthermore, the analysis of each parameter’s sign indicates that the observed behavior aligns with expectations, providing meaningful insights into the factors influencing decision-creating.

More specifically, passenger safety emerges as the most decisive factor shaping European citizens’ attitudes toward road automation. The model reveals that ensuring passenger safety has a substantial positive effect on utility, confirming its critical role. This aligns with earlier statistical analyses and underscores safety as a non-nereceivediable priority for respondents. Regarding data privacy, the findings confirm that the absence of robust state and indusattempt guarantees for data protection discourages the apply of autonomous vehicles. Similarly, the legislative framework significantly influences decision-creating; respondents favor adopting new technology only when the authorities ensure there are no legislative gaps or uncertainties. For road infrastructure, respondents are positively influenced by the existence of necessary infrastructure improvements that enhance passenger safety. Economic considerations also play a role. Respondents with a monthly income exceeding 5,000€ are more inclined to prefer an autonomous vehicle for their transport compared to those with lower income levels.

In addition to the above five key variables, several other statistically significant coefficients offer applyful insights. For instance, individuals who are unaware of the presence of autonomous public transport systems in Europe display a lower likelihood of adopting AVs, underscoring the role of information dissemination. Female respondents exhibit a slightly higher inclination to adopt AVs, potentially reflecting greater sensitivity to safety and mobility benefits. Furthermore, individuals who travel primarily for entertainment purposes display increased interest in AV adoption, suggesting that trip purpose influences openness to new mobility technologies.

Additional model outcomes: Number of observations: 4700

Number of individuals: 235

Null loglikelihood (with zero coefficients): -3129.75

Loglikelihood at convergence: -2966.50

Number of Fisher Scoring iterations (to converge): 4

Moreover, a positive attitude towards new technology and prior familiarity with autonomous vehicles are significant factors. Individuals who either possess knowledge of AVs, have experience driving highly automated vehicles, or express greater trust in them compared to conventional vehicles, exhibit a strong positive influence on adoption. Conversely, a lack of awareness about existing autonomous public transport in Europe has a negative impact on decisions related to autonomous vehicle apply.

Finally, socio-demographic characteristics further refined the analysis. Respondents aged 36–45 demonstrate a favorable inclination toward applying autonomous vehicles, comparing with those over 65 years old, who are less supportive. Commuting habits also provide valuable insights. Respondents who primarily commute via motorcycles or cars as drivers exhibit a positive attitude toward the new technology. Similarly, those whose primary reason for commuting is entertainment express notable support for autonomous vehicles. These findings collectively highlight the multifaceted considerations influencing public adoption of autonomous vehicles, emphasizing the importance of safety, infrastructure, legal certainty, economic factors, and socio-demographic attributes.

These findings reinforce and extfinish existing research on the factors shaping AV acceptance. For example, the significance of safety, trust in automation, and regulatory clarity is consistent with previous work by Chen et al. (2022)⁴⁶, Kaplan et al. (2019)⁴⁵, and Al Mansoori et al. (2024)³⁸, who identified these elements as essential for fostering public confidence in AVs. Our results further validate the significant influence of demographic factors such as income and age, aligning with the observations of Wang & Zhao (2019)³⁹ and Alawadhi et al. (2020)⁴⁷. Additionally, they confirm that legal uncertainty and concerns over data privacy continue to pose major obstacles, as highlighted by Raj et al. (2020)⁴⁸. Importantly, this study contributes new perspectives by highlighting the role of information awareness and trip purpose, as significant determinants of AV adoption. For example, individuals traveling for leisure or those unfamiliar with existing AV services exhibit distinct attitudes, suggesting that exposure and communication strategies may strongly shape public readiness. Collectively, these findings highlight the importance of a more nuanced, context-specific understanding of AV acceptance in Europe and emphasize the necessary for comprehensive policy approaches that tackle both technological challenges and societal perceptions.

Source link