Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Diese Studie untersucht die sozioökonomischen Faktoren, die das Reiseverhalten in der Schweiz beeinflussen, und nutzt hochauflösende GPS-Trackingdaten aus dem Projekt TimeUse +. Die Studie untersucht drei Schlüsselaspekte: die tägliche Fahrtfrequenz, die tägliche Reisezeit und die Wahl der Verkehrsart. Dabei werden unterschiedliche Muster zwischen Wochentagen und Wochenenden sowie zwischen Personen mit und ohne Autozugang aufgezeigt. Die Analyse unterstreicht den Einfluss von Geschlecht, Alter, Bildung, Beruf und Haushaltsmerkmalen auf das Reiseverhalten. Bemerkenswert ist der Studie zufolge, dass Frauen aufgrund ihrer Rolle als Betreuerinnen weniger Reisen unternehmen, während ältere Erwachsene eine höhere Reisehäufigkeit und längere Reisezeiten aufweisen. Der Besitz von Autos beeinflusst das Reiseverhalten erheblich, wobei Autobesitzer ihre Fahrzeuge seltener, aber für längere Fahrten nutzen, während Nichtautobesitzer stärker auf öffentliche Verkehrsmittel und aktive Verkehrsmittel angewiesen sind. Die Studie unterstreicht auch die Rolle, die Abonnements des öffentlichen Nahverkehrs bei der Gestaltung des Reiseverhaltens spielen, wobei Monatstickets zu einer höheren Nutzung unter Pendlern führen. Die Ergebnisse deuten darauf hin, dass eine Politik, die auf Autobesitzer und Nicht-Autobesitzer abzielt, eine nachhaltigere und gerechtere Mobilität fördern könnte. Die Studie kommt zu dem Schluss, dass sozioökonomische und räumliche Kontextfaktoren zwar wichtig sind, aber auch andere Faktoren wie Wetterbedingungen und disruptive Ereignisse eine bedeutende Rolle im Reiseverhalten spielen. Die Forschung betont die Notwendigkeit umfassender Studien, die diese Faktoren integrieren, um ein vollständiges Bild des Mobilitätsverhaltens zu liefern.
KI-Generiert
Diese Zusammenfassung des Fachinhalts wurde mit Hilfe von KI generiert.
Abstract
Understanding the relationship between socio-economic factors and travel behavior is crucial for developing sustainable and inclusive transportation policies. This study analyzes mode choice, daily trip frequency, and daily travel time in Switzerland using high-resolution GPS data from the TimeUse+ project, which provides detailed mobility records from over 1,000 individuals. Unlike traditional survey-based studies, GPS tracking offers more precise insights into daily mobility patterns by reducing recall bias and capturing short trips often omitted in self-reported data. Our analysis reveals clear differences between weekday and weekend travel: On weekdays, socio-demographic variables (such as gender and age) together with transport supply variables (such as car ownership and public transport subscriptions) significantly shape mobility while on weekends, these effects largely weaken, and lifestyle-driven choices dominate. Notably, adults aged 66 + are the most mobile group on weekdays in terms of trip frequency, while individuals aged 56–65 take fewer but longer trips. Women make fewer trips overall, likely due to higher load of household-related responsibilities. Households with children reduce travel time but not trip frequency highlighting the complex effects of caregiving roles. Car access remains the strongest determinant of mode choice, with lower-income car owners showing higher car dependency than higher-income ones. Public transport subscriptions—especially monthly passes—strongly increase weekday transit use. On weekends, older adults and women show greater openness to sustainable modes like walking and cycling, indicating untapped potential for non-car mobility.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
Understanding the relationship between socio-economic factors and travel behavior is critical for developing sustainable, efficient, and inclusive transportation systems. In Switzerland, a country renowned for its high-quality public transportation infrastructure and significant active mobility rates, the determinants of mode choice, daily trip frequency (number of daily trips), and daily travel time reflect a complex interplay of individual circumstances, spatial context, and policy frameworks. This study aims to explore these relationships by leveraging high-resolution GPS tracking data from the TimeUse+ project (2020), which provides detailed, app-based records of daily trips for more than 1,000 individuals over extended periods.
Most studies on travel behavior rely on survey-based methods, which, while widely used, present several limitations as input data for analyzing mobility patterns. Travel surveys are prone to recall bias, inaccuracies in self-reported trip details, and limited temporal and spatial granularity (Bhat and Misra 1999; Limtanakool et al. 2006). These limitations can affect the reliability of findings, especially when studying detailed mode choice behavior. In contrast, GPS tracking data offer objective, continuous, and precise records of daily trips, capturing temporal patterns, spatial trajectories, and actual transport mode usage with higher accuracy. Unlike self-reported surveys, GPS-based tracking eliminates reliance on memory, reducing trip duration, frequency, and distance estimation errors. The finer granularity of GPS data also allows for a more nuanced analysis of individual and group-level travel behaviors. Nevertheless, GPS tracking also presents certain limitations: it does not capture socio-demographic characteristics, relies on algorithms to infer attributes such as mode choice and trip purpose, and may face continuity issues related to device usage or signal loss. These challenges underline the importance of combining GPS tracking with survey data and careful data validation—an approach that represents a key innovation of this study.
Anzeige
By integrating GPS data with socio-economic variables such as income, education, occupation, household size, and car ownership, this study examines how these factors influence daily trip frequency, travel time, and transport mode selection. This approach enables a more detailed understanding of travel behavior beyond aggregate trends. While GPS data provide superior accuracy and completeness, it is important to acknowledge that survey-based studies remain valuable for capturing subjective travel experiences, perceptions, and motivations—elements not directly recorded through GPS tracking.
The Swiss context provides an optimal setting for this research. The country’s commitment to sustainable mobility is reflected in policies promoting public transit, walking, and cycling, as well as efforts to reduce car dependency in urban areas. However, challenges such as rising urbanization, increasing environmental concerns, and disparities in mobility access persist. By analyzing how socio-economic factors affect mode choice daily trip frequency and daily travel time, this study contributes to addressing these challenges, offering evidence-based insights to support equitable and effective policy interventions. For example, our findings can help design mobility incentives tailored to specific population groups, and identify areas where access to transport is unequal, ensuring policies are more equitable and effective.
Background
Research on transport mode choice and travel behaviour has a long-standing tradition, with early work primarily relying on household travel surveys and stated preference surveys. These studies, such as those by Bhat and Misra (1999) and Limtanakool et al. (2006), focused on socio-economic factors like income, household composition, and car ownership to explain mode choice. While these methods have contributed valuable insights, they have limitations in capturing the dynamic nature of travel behavior and often suffer from biases due to recall errors and incomplete trip records. More recent advances have shifted toward GPS tracking data, which enables the capture of precise, real-time data on travel behavior, eliminating many of the biases inherent in self-reported surveys.
The integration of GPS data into mobility studies has revolutionized our ability to analyze mode choice and travel behaviour at a granular level. Notably, Montini et al. (2017) demonstrated how GPS data can provide a more accurate picture of urban mobility, showing how different socio-economic groups engage with transportation systems in terms of travel behaviour and mode choice. Similarly, Crawford (2020) explored the spatial–temporal patterns of commuters’ daily trips, revealing differences in travel behavior based on sociodemographic factors, like age, gender, travel costs, and bike availability. By leveraging large-scale GPS datasets, these studies have highlighted the potential for more data-driven, individualized models of travel behavior.
Anzeige
Another influential body of work has explored the intersection of spatial context and mobility behavior. Studies like those by Chen et al. (2008), Harbering and Schlüter (2020), and Schlüter et al. (2021) underscore the importance of integrating environmental factors—such as urban density, access to public transport, and proximity to key destinations—into models of mode choice. These studies argue that individual travel behavior cannot be understood in isolation from its spatial context, suggesting that a more nuanced analysis of socio-economic factors requires an understanding of both the built environment and the individual’s personal circumstances. This aligns with the spatial economics approach, which has been applied in transportation studies to examine how place-based factors interact with socio-economic variables to shape mobility decisions (see Banister 2008, and Graham and Marvin 2001).
In addition, transportation behavior segmentation—the process of identifying different groups within the population with similar travel patterns—has become an important research direction. Lee et al. (2019) and Wang and Shen (2022) focused on latent class analysis and other statistical techniques to uncover distinct mobility behaviors within different socio-economic cohorts. This approach is particularly relevant when investigating the mobility choices of specific groups, such as low-income households or seniors, who may face barriers to accessing conventional modes of transportation, thus altering their patterns of travel.
More recently, app-based studies such as Moovit (2012) and TimeUse+ (2020) have taken a significant step forward by integrating real-time smartphone-based tracking with socio-economic profiles. These studies capture both the fine-grained temporal and spatial data of travel and the individual-specific socio-economic characteristics that influence travel behavior. The TimeUse+ project is a prime example of this innovation, allowing researchers to study daily trips in a more comprehensive and dynamic way, enabling the integration of individual-level mobility data with socio-economic indicators like household composition, education level, and vehicle ownership. Moovit, in particular, has been widely studied for its impact on urban mobility, demonstrating how app-based real-time transport information can increase public transport usage and contribute to Mobility as a Service (MaaS) solutions (González-Sánchez et al. 2024). These datasets are crucial in revealing the role of personal circumstances in determining travel patterns, offering insights into how different segments of the population make daily travel decisions.
Most studies on mode choice continue to rely on regression-based approaches such as multinomial logit, mixed logit, or hierarchical models to estimate how socio-economic and contextual factors influence mobility decisions. These methods remain widely used due to their interpretability and relevance for transport policy analysis. While machine learning techniques like random forests and neural networks have gained attention for their ability to model complex and non-linear relationships, their lack of transparency and focus on prediction rather than explanation often limits their applicability for studies aiming to uncover underlying behavioral mechanisms.
Despite advances in GPS-based travel studies, few have fully integrated high-resolution, individual-level mobility data with detailed socio-economic profiles to investigate daily mobility metrics beyond mode choice, such as trip frequency and daily travel time. This study addresses this gap by leveraging the TimeUse+ dataset, which combines app-based GPS tracking with rich socio-demographic data for more than 1000 individuals. Rather than applying novel machine learning models, this research reexamines foundational questions in travel behavior using higher-quality data, offering more precise and empirically grounded insights into how socio-economic characteristics shape everyday mobility beyond mode choice.
In addition, while previous studies have made significant strides in understanding various aspects of travel behavior, they often focus on individual components such as mode choice, trip frequency, activity duration, trip distance and other travel metrics, without fully integrating these dimensions into a broader analysis of daily mobility. Research like Li et al. (2021) has examined the influence of socio-economic factors on activity duration, but this is limited to specific activities, not travel as a whole. Studies by Xu and Wang (2018) and Singh et al. (2024) explore trip frequency, but they typically rely on travel surveys, which provide less granular data compared to high-resolution GPS tracking. Similarly, Lee et al. (2023) analyzed changes in daily trip frequency and trip distance between public and motorized transport modes during the COVID-19 pandemic, using data from public transport and traffic counters. However, the use of counters provides an approximation of trip distance, which lacks the precision and reliability that GPS tracking data offers for measuring exact travel distances. Additionally, Benita (2023) explored how trip frequency is associated with different transport modes but again relied on travel surveys. These studies, while valuable, tend to concentrate on specific mobility metrics rather than providing an integrated view of daily travel behavior across modes. Moreover, their reliance on travel surveys limits the precision and accuracy of the findings.
Our research addresses this gap by leveraging high-resolution GPS tracking data combined with socio-demographic information to analyze total daily travel time, trip frequency, and mode choice. Unlike previous studies that focus on one or two aspects of mobility, we offer a holistic analysis of daily mobility patterns, accounting for the cumulative travel behavior across an entire day. By doing so, we provide a more accurate and nuanced understanding of how socio-economic factors influence individual travel behavior. This approach allows us to bridge the gap between mode choice studies and broader mobility metrics, offering new insights into the complexities of everyday transportation.
Data
Data source (TimeUse+)
Data collection for the main study commenced in July 2022, with 3000 potential participants invited each week until December 2022. Participants were required to take part for a duration of four weeks, extending the overall study period through February 2023. The study was structured into three phases: an initial questionnaire, a four-week tracking and validation period, and a final questionnaire following the tracking phase. A total of 1318 individuals successfully completed all study components, including the four-week travel diary and associated online surveys. The study was conducted only in a German-speaking part of Switzerland (Winkler et al. 2024).
Variables
Dependent
To model socio-economic and spatial context effects on travel mode choice number of daily trips and daily travel time, we defined three dependent variables, one for each model. The dependent variable travel mode is composed of five different values: walking, bike, public transport, motorized private transport (mpt), and train. Another two variables: the number of daily trips and daily travel duration are modelled as the normalized value using min–max normalization (i.e., rescaling each value between 0 and 1 based on the observed minimum and maximum) of daily trips per individual and the total duration of daily trips per individual, respectively. To address the presence of extreme daily trip and travel time values which seem to be very unrealistic (e.g. 49 daily trips or 15 h of daily travel) we applied interquartile range (IQR) filtering (Tukey 1977). Q1 is the first quartile (25th percentile), Q3 is the third quartile (75th percentile), and IQR = Q3 − Q1. The lower and upper bounds for both variable’s values filtering were calculated using the equations provided in Eqs. 1 and 2.
$$lower\_bound = Q1 - 1.5 * IQR $$
(1)
$$upper\_bound = Q3 + 1.5 * IQR $$
(2)
where Q1 is the first quartile, Q3 is a third quartile and IQR is an interquartile range.
Following this process, the maximum allowable number of daily trips was set to 19 for weekdays and 18 for weekends. Similarly, the maximum allowable daily travel duration was set to 8.1 h for weekdays and 8.5 h for weekends. Consequently, the normalization of the dependent variables was performed based on these adjusted maximum values (‘min–max normalisation’).
Independent
The independent variables include travel cost, travel distance, and a range of socio-economic variables. These socio-economic variables are: gender, age group, education, occupation, driving license, car access, public transport subscription, bicycle access, residential location area, number of kids, and monthly household income. As we can see, only two of them are continuous variables, whereas the others are categorical ones. Table 1 in presents categorical variables and their distribution in the dataset.
Table 1
Categorical variables specification and distribution
Variable
Category
Distribution (%)
Gender
Male
54
Female
46
Age Group
(18–40)
(41–55)
(56–65)
(66 +)
37
35
18
10
Education Level
Higher education (e.g., university)
47
Mandatory education
2
Secondary education (e.g., apprenticeship or diploma)
51
Occupation
Employed
82
Retired
9
Student
4
Unemployed
5
Bike Access
No
16
Yes
84
Car Access
No
27
Yes
73
Public Transport Subscription
GA
14
No
22
Other
52
Monthly ticket
13
Trip Purpose
Errands
2
Home
35
Leasure
13
Other
19
Shopping
8
Work
23
Household Location Type
City center
6
Rural
47
Suburban
30
Urban
17
Household young kids (younger than 13)
FALSE
74
TRUE
26
Household income
4000 CHF or less
10
4001–8000 CHF
32
8001–12,000 CHF
35
12,001–16,000 CHF
14
More than 16,000 CHF
9
Household_size
1
16
2
37
3
17
4
24
5 or more
6
The dataset consists of a balanced gender distribution (54% male, 46% female) and a diverse age range, with the majority (73%) falling between 18 and 55 years old. Most participants have secondary (51%) or higher education (47%), and employment is the dominant occupation (82%). Bike access is high (84%), while public transport subscriptions vary, with a notable share categorized as ‘other’ (52%). Regarding trip purposes, home-related activities dominate (35%) meaning trips with home as either the origin or destination, while work and leisure trips account for 23% and 13%, respectively. Household characteristics show a mix of urban and rural living, with 47% in rural areas. Household income and savings distributions indicate variability, with the largest group earning between 8001 and 12,000 CHF (35%). The dataset includes a diverse range of socio-economic and household characteristics, ensuring a representative sample for mode choice analysis. The distribution of key attributes such as occupation, income, and transport access provide sufficient variation to capture different mobility patterns. Continuous variables and their statistics are presented in Table 2.
Table 2
Continuous variables statistics
Variable
Mean
Median
Std. Dev
Trip cost
3.10 CHF
1.10 CHF
7.49 CHF
Trip distance
19.87 km
5.82 km
42.55 km
The values from the Table 2 indicate a highly skewed distribution for both variables. The mean trip cost (3.10 CHF) is significantly higher than the median (1.10 CHF), suggesting that a small number of expensive trips increase the average. Similarly, the mean trip distance (19.87 km) is much larger than the median (5.82 km), indicating that most trips are relatively short, but a few very long trips pull the average upward. The high standard deviations (7.49 CHF for cost and 42.55 km for distance) further confirm the presence of large variations.
Descriptive statistics
This sub-section provides a deeper overview of some of the factors that are related to mode choice.
Figure 1 depicts different modal splits indicating the choice of travel modes by gender.
Fig. 1
Percentage share of trips by travel mode depending on gender
Both genders follow the same overall trend in mode choice, but some differences exist in their relative shares. Men have a higher share of motorized private transport—mpt (44% vs. 37%), suggesting a greater reliance on private vehicles. Women, on the other hand, show slightly higher usage of public transport-local_pt (local public transport and train combined: 30% vs. 26%) and active modes such as walking (26% vs. 23%). Biking remains the same for both genders (7%). These differences, though not extreme, suggest that women may be slightly more inclined toward public and active transportation, while men rely more on private vehicles. Figure 2 depicts different modal splits indicating the choice of travel modes by participant’s occupation.
Fig. 2
The percentage share of trips by travel mode depending on occupation
The travel behavior patterns across employment statuses reveal some important similarities and distinctions. Both unemployed and retired individuals show similar preferences, with walking being the most common mode of transport (28% and 29%, respectively), followed by motorized private transport (41% for retirees and 39% for the unemployed). These similarities may reflect a more localized lifestyle with fewer long-distance commuting needs. On the other hand, employed individuals display a significantly higher preference for motorized private transport (43%) and train travel (17%), likely due to regular work-related travel. Students differ significantly from other groups in their travel behavior. They rely heavily on local public transport (25%) and trains (36%), possibly due to more affordable or subsidized travel options and the typically shorter commuting distances. This stands in contrast to employed, unemployed, and retired individuals, who demonstrate a stronger reliance on mpt. Students also use mpt much less, with only 13% opting for it, likely reflecting a combination of financial constraints and the availability of more convenient public transportation options for their daily trips.
The travel behavior patterns across different home locations reveal interesting distinctions based on proximity to urban centers and available transport options (see also Fig. 3). Individuals living in city centers tend to walk the most (33%) and use public transport (16%), indicating the availability of convenient and accessible transport infrastructure. They also have a relatively moderate use of mpt at 28%. In contrast, those residing in urban areas show a slightly more balanced distribution of transport modes, with a stronger preference for train travel (21%) and slightly higher mpt usage (25%), but still maintaining a significant reliance on walking (27%). Residents of suburban and rural areas show more reliance on motorized transport, with suburban residents using motorized private transport the most (40%) and rural residents even higher at 51%. This suggests that as the distance from the city center increases, the use of mpt becomes more pronounced due to less access to frequent public transport and walking infrastructure. In these areas, walking is still a common mode but represents a smaller proportion of the overall travel behavior, with rural areas showing the lowest walking percentage (23%). These differences underscore the influence of the built environment and available infrastructure on travel mode choices, with urban areas favouring more sustainable and accessible modes, while suburban and rural areas lean towards motorized transport due to the limitations of public transit.
Fig. 3
The percentage share of trips by travel mode depending on home location
The transport mode choices of car owners and non-car owners (see Fig. 4) reveal distinct patterns: non-car owners rely more on public transport, walking, and trains, reflecting a greater dependence on shared and active mobility options. In contrast, car owners show a strong preference for motorized private transport, with significantly lower usage of public transport and trains, suggesting a reliance on the convenience and flexibility of private vehicles. Both groups exhibit low bike usage, indicating potential barriers such as inadequate infrastructure or safety concerns. These differences highlight the influence of car ownership on travel behavior. However, self-selection mechanisms may also play a role: individuals may choose residential locations and transport subscriptions based on their preferences, which in turn affect car access and availability. As our dataset does not include information on individual preferences, we cannot separate these effects, and the results should therefore be interpreted in light of this limitation.
Fig. 4
The percentage share of trips by travel mode depending on car ownership
In this section, we explore three key associations relevant to our study. We analyze how socio-economic characteristics influence (1) mode choice, (2) the number of daily trips, and (3) daily travel time.
The first association will be modelled using multinomial logistic regression, which is particularly well-suited for mode choice analysis due to its ability to handle a categorical dependent variable such as travel mode choice. This approach is suitable for examining how various explanatory factors—such as socio-economic characteristics, trip cost, or distance—affect the probability of choosing one transportation mode over another. By estimating the relative odds of selecting each mode, multinomial logistic regression provides valuable insights into the underlying patterns of travel behavior. Furthermore, its \(\beta_{J}\) enhances its utility in understanding complex decision-making processes. The widespread use of this method in transportation research further reinforces its relevance and appropriateness for our study (Cervero and Kockelman 2004; Sener et al. 2010; Bhat et al. 2015; Zhang et al. 2023).
where (y = J|X) the probability of the outcome being in category j, \(\beta_{J}\) is the vector of coefficients for category \( j\) (relative to the reference category \(J\)), \(X\) is the vector of independent variables (predictors), \(P\left( {y = J|X} \right)\) is the probability of the outcome being in the reference category \(J\), which is not explicitly modeled but is instead used as the baseline.
The odds ratio for each predictor \(X\) in category \( j\) is given by the exponential of the coefficient \(\exp \left( {\beta_{J} } \right)\), which gives the multiplicative change in the odds of being in category \( j\) relative to the reference category \(J\).
Further, we use Generalized Linear Models (GLM) with a Gaussian distribution and identity link to model the influence of socio-economic factors on the number of daily trips (trip frequency) and total daily travel time. The dependent variables in both cases are the normalized values of the mentioned variables, which are continuous outcomes. Although the data for both outcomes does not follow a normal distribution (as indicated by the Shapiro–Wilk test, which suggests skewness in the data), the GLM with Gaussian distribution can still be effective for this purpose. The identity link function is applied, meaning the model predicts the expected value directly, allowing the model to handle the relationship between the predictors (socio-economic and spatial factors) and the outcomes in a linear fashion. The GLM with a Gaussian distribution and identity link function was used by Washington et al. (2003) to model the number of traffic accidents, utilizing various explanatory variables.
The reason this specific GLM is appropriate despite the data’s skewness is that the identity link function allows the model to predict the raw, untransformed outcome values. While skewness indicates non-normality, the Gaussian GLM with an identity link can still be robust in modeling outcomes when the skewness is not extreme. It assumes a linear relationship between predictors and the response, and it is flexible enough to model continuous, normalized data like trip frequency and travel time without requiring strict normality Manning and Mullahy (2001). This choice provides a detailed understanding of how different factors influence daily trips and travel time while accommodating for the skewed nature of the data.
The Gaussian GLM with an identity link is defined in the formula:
$$ g\left( \mu \right) = X\beta $$
(4)
where \(g\left( \mu \right) = \mu\) (identity link function, where μ is the expected value of the dependent variable Y), \(\mu = E\left[ {Y|X} \right]\) is the conditional mean of the dependent variable Y, \(X\) is the vector of independent variables (predictors), β is the vector of coefficients (parameters).
To better analyze mobility patterns, the dataset was partitioned into four subsets. The first partition is done into weekend and weekday trips. This distinction is justified by the markedly different travel behaviours observed between the two periods (Agarwal 2004; Ho and Mulley 2013; Orowo et al. 2022). Weekday trips are predominantly influenced by structured activities such as commuting and errands, which are constrained by time and shaped by socio-economic factors. In contrast, weekend trips tend to be associated with leisure and discretionary activities, characterized by different spatial and temporal dynamics. Further partitioning is applied when analyzing the impact of socio-economic factors on mode choice by distinguishing between travellers with and without car ownership. This is important because car ownership has a significant impact on travel behavior. People with a car are more likely to use it for both weekday and weekend trips, especially for longer trips or commuting. Those without a car, however, may heavily rely on public transport, walking, or biking, and this can affect how they travel for both work and leisure. We believe that this partition ensures that socio-economic factors affecting each group are better understood and analyzed independently, leading to more accurate and targeted insights into mobility patterns.
Therefore, to evaluate the impact of socio-economic factors on travel mode choice, we create four separate multinomial logistic regression models based on both trip timing and car ownership status. These models focus on mode choice for weekend and weekday trips, distinguishing between individuals with and without car access. Given that the dataset is already partitioned by car ownership, this variable is excluded from these models. In contrast, for our analyses of daily trip frequency and daily travel duration, the separation is done only between weekday and weekend while car ownership is included as a covariate because it plays an integral role in overall mobility patterns, influencing the number and length of trips across all individuals.
Results
Mode choice analyses
Our regression analyses of travel mode choice were conducted separately for weekdays and weekends and further stratified by car ownership.
This multinomial regression analysis of a mode choice among car owners for weekday trips (see Appendix 1, Table 7) show that gender plays a role, with women more inclined toward local public transport and less likely to choose motorized private transport compared to men. Age also matters—older individuals tend to prefer biking over walking, while the oldest age group shows a strong preference for trains, with 7 times greater odds of choosing trains over walking, which corresponds to a 7 times higher probability (likelihood) of doing so. Education has a significant impact, with higher education levels making motorized modes, especially local public transport, much more likely. Employment status further shapes choices, with employed individuals favouring motorized options, while students are far less likely to bike, but are 2 times more likely to use public transport over walking. At the same time retired individuals are 6 times more likely to choose a public transport instead of walking. Household characteristics, such as bike access and size, influence decisions—having a bike increases cycling, while larger households lean toward motorized modes. However, households with young children are less likely to choose any non-walking option. Public transport subscriptions strongly predict public transport use—particularly train use—with GA and monthly ticket holders being 19 and 7.7 times more likely respectively, to choose trains over walking, whereas higher income generally reduces the appeal of motorized transport. In this sense, PT subscription is not a purely exogenous factor, and our findings should be interpreted in light of this potential endogeneity. Trip-related factors also play a role: longer distances encourage biking and train use but discourage local public transport, while work and errand trips limit mode variety compared to leisure trips. Finally, urban residents, particularly those in city centers, prefer biking and local transit but avoid trains, while suburban dwellers show more balanced preferences.
Notably, the fact that these individuals are car owners adds an important dimension to these findings. Despite having access to a car, many still choose alternative modes—suggesting that car ownership does not necessarily equate to car dependency. Socioeconomic and urban factors, such as higher education and city-center living, appear to override the convenience of car use, pointing to a preference for efficiency, cost savings, or environmental consciousness in certain contexts. The strong influence of public transport subscriptions (e.g., GA travelcards) further indicates that convenience and financial incentives can outweigh car ownership when alternatives are viable. This behavior may reflect that car owners use their vehicles selectively—perhaps for specific trips (e.g., large shopping loads or bad weather) while relying on walking, biking, or transit for daily commutes. The lower preference for motorized modes during work or errand trips could imply that cars remain the default for utilitarian travel, but alternatives are competitive for other purposes. These findings challenge the assumption that car owners are resistant to sustainable transport. Instead, they suggest that policies targeting infrastructure (e.g., bike lanes), financial incentives (e.g., discounted transit passes), or urban design (e.g., congestion pricing) could further encourage mode shifts even among car-owning populations.
To visually summarize the model outcomes, Fig. 5 highlights the four most influential predictors of weekday mode choice among car owners. These predictors were selected based on their overall effect size across all transport alternatives. The plot displays odds ratios with confidence intervals for each predictor and mode, using a dot-whisker format to concisely convey both effect size and uncertainty. This visual complement supports the interpretation of model results presented in Appendix 1 and reinforces the dominant role of access-related and socio-demographic variables in shaping weekday mobility behavior. Odds ratios greater than 1 indicate a positive association with the mode, while values below 1 indicate a negative association.
Fig. 5
Top 4 most influential predictors of transport mode choice for weekday trips among car owners
The findings related to non-car owners’ weekdays mode choice (see Appendix 1, Table 8) reveal clear differences compared to car owners, shaped by necessity rather than preference. Women are less inclined toward motorized transport but more likely to take trains, while older adults, especially those over 66, demonstrate a strong preference for all other modes over walking, suggesting a strong tendency to choose alternatives rather than walk, probably due to their physical constraints. Unlike car owners, higher education reduces cycling but increases public transit use with 2 times greater likelihood over walking, suggesting that education correlates with transit reliance when car access is absent. This may partly reflect the location of highly qualified jobs that are more concentrated in dense urban environments with better public transport provision. Bike availability has a dramatic effect, significantly boosting cycling while reducing local transit use—a substitution effect not seen among car owners. Public transport subscriptions, particularly travel passes, have an outsized influence, making trains and public transport the default choice for many. Household structure and income also play distinct roles. Larger families lean toward motorized transport and trains, likely for practicality, while households with young children stick to walking more than car-owning families do. Interestingly, the positive and strong association between trip cost and public transport use in Zurich suggests that higher costs do not discourage usage for both car and non-car owners on weekdays. Higher-income non-car owners are more likely to use ride-hailing or taxis, whereas lower-income groups depend more on public transit. Trip characteristics further shape decisions: work-related trips increase train use, while errands and leisure trips favor walking. Longer distances push people toward biking and trains but discourage local transit, highlighting gaps in public transport coverage. Urban dwellers rely heavily on transit, while those in suburban and rural areas face limited options, sometimes resorting to motorized transport despite not owning a car.
The results suggest that non-car owners adapt their travel behavior based on available infrastructure and financial constraints. Their choices are more sensitive to incentives like bike lanes and transit subsidies, making them a key demographic for sustainable mobility policies. Unlike car owners, who may alternate between driving and other modes, non-car owners depend entirely on walking, biking, and public transit—meaning targeted improvements in these areas could significantly reduce reliance on motorized transport. Cities aiming to promote car-free living should focus on making alternatives more accessible, affordable, and convenient for this group. The low likelihood of cycling among both car owners and non-car owners in rural areas points to inadequate cycling infrastructure as a key deterrent.
As illustrated in Fig. 6, weekday travel patterns among non-car owners are primarily influenced by public transport subscription status, and employment factors. The figure summarizes the model’s strongest effects.
Fig. 6
Top 4 most influential predictors of transport mode choice for weekday trips among non-car owners
Weekend travel patterns for car owners (see Appendix 1, Table 9) reveal significant shifts from weekday routines, particularly in mode preferences and trip purposes. While weekdays showed stronger reliance on motorized transport for work and errands, weekends feature increased biking among higher-educated individuals (exceptionally high odds ratios) and those with bike access, suggesting recreational cycling dominates. Train use also gains importance with high odds, especially among older adults (66+ age group), likely for leisure trips, contrasting with their weekday avoidance of motorized modes. However, employed individuals and families with young children remain less likely to choose active modes, indicating persistent car reliance for family-oriented weekend activities. Comparing weekend and weekday patterns reveals key behavioral differences. Bike access has a significantly stronger influence on mode choice among car owners during weekends compared to weekdays. Public transport subscriptions maintain influence but operate differently—while weekday commuting reinforced local transit use, weekends see train travel surge (especially with GA passes), yet local transit becomes less preferred. Income effects reverse somewhat, with higher-income households avoiding trains more on weekends than weekdays, possibly opting for car-based leisure trips. Urban residents continue using local transit twice more than rural counterparts, but the urban–rural divide softens slightly as even city dwellers reduce transit use for non-work trips. The bigger households such as 5 + members are 3 times more likely to use trains than to walk. High trips costs are strongly associated to cars, train an public transport during the weekend showing the willingness to prioritize weekend spare time activities over costs.
These shifts highlight how car owners adapt mode choices when freed from work obligations. Weekends unlock recreational travel patterns where biking and intercity trains gain prominence, while local transit becomes less essential. However, structural factors like family needs, income levels, and residential location still constrain true mode diversification. The persistence of car reliance for certain groups (families, high-income households) even on weekends suggests that overcoming habitual car use requires addressing these specific barriers through targeted incentives like family-friendly bike solutions or leisure-oriented transit packages.
The four strongest predictors for weekend travel behavior among car owners are shown in Fig. 7.
Fig. 7
Top 4 most influential predictors of transport mode choice for weekend trips among car owners
Weekend travel patterns among non-car owners (see Appendix 1, Table 10) demonstrate strong dependence on existing mobility infrastructure, with public transport maintaining its central role. Train usage shows particularly high association with subscription passes, while local transit sees reduced uptake compared to weekdays. Association of transport passes, and use of public transport and train is expectedly much stronger than in case of car owners. Bicycling emerges as a significant alternative for specific demographic groups, particularly those with direct bike access and higher education levels, though adoption varies substantially across age and gender lines. The analysis reveals clear demographic segmentation in mode choices. Women show consistently lower utilization across all transport options compared to men, while middle-aged adults exhibit greater versatility in mode selection. Household characteristics prove particularly influential, with larger family units facing substantially constrained mobility options and families with young children displaying distinct travel patterns. Urban versus suburban residence continues to shape travel behavior, though not as dramatically as might be expected, suggesting some level of service parity across locations. Urban and city-centre dwellers show strong likelihood of choosing walking among other modes showing they have a very localized activities during weekend. This finding suggests that, on weekends, Zurich is successfully realizing the concept of a 15-min city. Transport access variables demonstrate the most pronounced effects on weekend mobility. Bike availability strongly predicts cycling adoption, while public transport subscriptions overwhelmingly dictate train usage patterns. However, these relationships may also reflect self-selection, as individuals with strong preferences for cycling or public transport are more likely to maintain a bicycle or purchase a subscription in the first place. In this sense, access variables capture both provision and preference effects. Interestingly, income levels show minimal direct impact on mode choices, with subscription access appearing more influential than raw financial capacity. Trip distance strongly discourages local transit use for longer trips while having minimal impact on other modes, revealing that non-car owners maintain consistent preferences for biking, trains, and motorized transport regardless of journey length. Trip costs followed a similar pattern as observed among car owners during weekends.
Figure 8 shows the dominant factors influencing weekend travel among non-car owners. The visual reinforces the importance of access to public.
Fig. 8
Top 4 most influential predictors of transport mode choice for weekend trips among non-car owners
The presented visualizations (see Figs. 5, 6, 7, 8) showed that while many predictors are statistically significant across the models, it is the most influential ones (those with the largest effect sizes) that show notable consistency. Among car owners, the four most influential predictors remain exactly the same between weekday and weekend models. For non-car owners, there is also strong alignment, with three out of the top four predictors overlapping. This suggests that, despite variations in significance across the full set of variables, a core group of predictors—mainly related to transport access and socio-demographic characteristics—consistently plays a dominant role in shaping mode choice across different temporal contexts.
Beyond these top predictors, the full set of results shows that travel behavior reflects both structural constraints (car access, income, geography) and temporal rhythms (weekday necessity vs. weekend agency). As preferences are not captured in our dataset, our results do not cover the full range of influences on mode choice, but instead focus on the observable effects of socio-economic characteristics and transport supply. While car owners remain tethered to vehicles, their weekend behavior reveals latent potential for sustainable shifts. Non-car owners, however, face systemic barriers that demand targeted redress. Policymakers must adopt integrated strategies that pair car-use disincentives with infrastructure upgrades and equity-driven subsidies, ensuring mobility systems are both sustainable and inclusive. In addition, addressing geographic inequities requires tailored strategies: urban areas need bike lane expansions, suburbs require integrated rail-bus networks, and rural regions benefit from on-demand shared transport pilots.
Number of daily trips (daily trip frequency)
Following the approach proposed in the previous section we first applied GLM regression models with Gaussian distribution and identity link to address socio-economic factors effects on number of daily trips. Table 3 presents results of regression model run on weekday dataset.
Table 3
Effects of socio-economic factors on the number of daily trips – weekday travel pattern
Education [Secondary education (e.g., apprenticeship or diploma)]
− 0.0003
0.003
Occupation [retired] ref. (employed)
− 0.0788 ***
0.014
Occupation [student]
0.0293 ***
0.008
Occupation [unemployed]
− 0.0164 *
0.007
Driverslicense [Yes] ref. (No)
0.0202 **
0.007
car_access [Yes] ref. (No)
− 0.0094 *
0.004
bike_access [Yes] ref. (No)
0.0109 **
0.004
public_transport_subscription [NO] ref. (GA)
− 0.0776 ***
0.005
public_transport_subscription [OTHER]
− 0.0770 ***
0.005
public_transport_subscription [monthly ticket]
0.0039
0.006
hh_loc_type [Rural] ref.(city_center)
0.0034
0.007
hh_loc_type [Suburban]
0.0176 **
0.007
hh_loc_type [Urban]
0.0039
0.007
hh_income [4000 CHF or less] ref. (12,001–16,000 CHF)
0.0060
0.006
hh_income [4001–8000 CHF]
0.0026
0.005
hh_income [8001–2000 CHF]
0.0175 ***
0.004
hh_income [More than 16,000 CHF]
0.0156 **
0.006
hh_young_kids [True] ref (False)
− 0.0059
0.003
N (Number of observations): 24,386
LLnull (Log-Likelihood of the null model): 2058.5723248044637
LLfinal (Log-Likelihood of the fitted model): 2519.521962677406
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
The analysis of the (total) number of daily trips per day, based on a weekday mobility pattern, reveals significant effects of several socio-economic and demographic factors. Gender differences are evident, with women making fewer daily trips compared to men, which aligns with findings in previous transport studies indicating that women often engage in more localized, shorter trips primarily related to caregiving or household duties. Age also plays a substantial role, older age groups, specifically those between 41 and 65, tend to make fewer trips compared to the reference group 18 to 40. However, individuals aged 66 and older display an increase of 7.9% in trip frequency, which may reflect a higher need for leisure and health-related travel. Additionally, many in this age group may still be professionally active, often working part-time which is a common scenario in Switzerland. These results underline the importance of considering age as a determinant in transportation planning, particularly when targeting policies for specific age groups.
Occupation and income influence daily trip frequency, with retired and unemployed individuals making fewer trips and students exhibiting a higher frequency of travel than employed ones, while higher income levels are associated with more trips, but lower-income groups do not show significant differences. This also shows that the 66+ age group is divided into two main segments: one that is retired and another that remains professionally active. The differing behaviours between these groups could be further influenced by early retirements due to personal choice or disability. Household income is positively correlated with the number of daily trips, especially among those earning higher income brackets, which may reflect greater access to transportation options, as well as the potential for more frequent travel due to work or personal commitments. The positive association between having a driver’s license and daily trips suggests that those with a license may feel more independent and likely to take trips, even if they do not own a car, potentially using alternatives like public transport or cycling. This effect may also reflect overlapping influences of license-holding and car access, which are conceptually related and not fully separable in our analysis. By contrast, the negative correlation between car access and daily trips may indicate that people with a car are more selective about when to use it, consolidating and optimizing trips due to associated costs and convenience. In contrast, individuals with access to a bicycle tend to make more trips, suggesting that cycling is an important mode of transport during weekday routines. Bicycles are broadly used for various activities, with many people depending more on their bikes than on cars for commuting and daily tasks. This shift highlights how cycling has become a preferred and versatile mode of transport, particularly in urban settings where it offers flexibility and convenience.
People residing in suburban areas are more active in terms of daily trips compared to those in urban or rural areas, potentially due to the spatial distribution of services and the necessity of travel for accessing amenities in less dense areas. Public transport subscription status also plays an important role, with individuals who do not subscribe to public transport or rely on alternative methods, such as pay-per-use services, making 7.8% less daily trips than those with GA. The presence of young children within a household surprisingly does not influence a number of daily trips significantly, probably due to the well-structured daily routines of parents. These findings emphasize the complex interaction between individual characteristics, access to transportation, and spatial context in determining weekday mobility patterns.
While several socio-economic variables show statistically significant associations with weekday trip frequency, the effect sizes are overall modest. For example hh_loc_type [Suburban], which illustrates that daily trip frequency remains relatively stable across socio-demographic groups. This suggests that significance reflects consistent but small differences rather than large substantive shifts in the number of daily trips.
Table 4 depicts the results of the GLM regression model with Gaussian distribution and identity link run on a weekend dataset.
Table 4
Effects of socio-economic factors on number of daily trips – weekend travel pattern
Education [Secondary education (e.g., apprenticeship or diploma)]
− 0.0112 *
0.005
Occupation [retired] ref. (employed)
− 0.0996 ***
0.024
Occupation [student]
− 0.0065
0.014
Occupation [unemployed]
− 0.0087
0.011
driverslicense [Yes] ref. (No)
− 0.0106
0.013
car_access [Yes] ref. (No)
− 0.0264 ***
0.007
bike_access [Yes] ref. (No)
0.0011
0.007
public_transport_subscription [NO] ref. (GA)
− 0.0317 ***
0.009
public_transport_subscription [OTHER]
− 0.0192 *
0.008
public_transport_subscription [monthly ticket]
− 0.0086
0.010
hh_loc_type [Rural] ref.(city_center)
− 0.0185
0.011
hh_loc_type [Suburban]
− 0.0181
0.011
hh_loc_type [Urban]
− 0.0001
0.011
hh_income [4000 CHF or less] ref. (12,001–16,000 CHF)
0.0236 *
0.011
hh_income [4001–8000 CHF]
− 0.0066
0.008
hh_income [8001–12,000 CHF]
0.0080
0.007
hh_income [More than 16,000 CHF]
0.0210 *
0.010
hh_young_kids [True] ref (False)
− 0.0026
0.006
N (Number of observations): 8903
LLnull (Log-Likelihood of the null model): 637.6610919454641
LLfinal (Log-Likelihood of the fitted model): 726.6236557393404
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
The analysis of weekend mobility patterns reveals a distinct shift in travel behavior compared to weekdays. While some socio-demographic factors remain influential, overall, there are fewer significant predictors of trip frequency on weekends. Gender differences persist, with women making slightly fewer trips than men. Older individuals, particularly those in the oldest age group, tend to make more trips on weekends (7% more than a reference group up to 40 years old). Education level plays a role, as those with lower educational attainment are generally less mobile probably due to the lower leisure time budget. Further, age-related differences and employment status have a weaker impact on weekend mobility compared to weekday patterns. Car access continues to be associated with fewer daily trips. While this may appear counterintuitive, it can be explained by the fact that car users often combine several purposes within a single journey (trip chaining), thereby reducing the number of trips recorded. In the TimeUse+ dataset, a trip is defined algorithmically as a movement between two stationary activity locations (‘stays’). Continuous GPS traces are segmented into stages of movement between stays, which are then aggregated into trips, with trip purpose assigned based on the destination activity (Winkler et al. 2024). This definition differs from declarative travel surveys, where each purpose is usually reported as a separate trip, which may explain why car owners appear with fewer but longer trips in our data. Public transport usage also influences weekend mobility, as those without a subscription travel for 3% less frequent than a reference group with GA passes. The magnitude of this effect was almost halved compared to the workdays. Interestingly, the effect of monthly public transport tickets, which is important on weekdays, disappears on weekends. This suggests that regular commuters rely less on public transport for discretionary weekend travel. Household location does not significantly shape weekend mobility, contrasting with its more pronounced role on weekdays, likely due to fewer obligations on weekends and greater freedom in choosing travel destinations. Income effect showed an interesting finding showing that lowest and highest income individuals are more active on weekend that medium income class. The smaller number of significant results compared to weekdays suggests that travel patterns on weekends are less structured and less influenced by socio-economic constraints. Unlike weekdays, when commuting and routine activities dictate travel frequency, weekend trips are often discretionary, varying by personal preferences, social activities, and leisure opportunities. This greater flexibility in weekend travel behavior results in weaker associations between individual characteristics and trip frequency.
Daily travel time
We also applied GLM with a Gaussian distribution and identity link model to address socio-economic factors effects on daily travel time. Table 5 presents results of regression model run on weekday dataset.
Table 5
Effects of socio-economic factors on extended travel duration—weekday travel pattern
Education [Secondary education (e.g., apprenticeship or diploma)]
− 0.0056
0.003
Occupation [retired] ref. (employed)
− 0.0204
0.015
Occupation [student]
0.0405***
0.009
Occupation [unemployed]
− 0.0114
0.007
driverslicense [Yes] ref. (No)
0.0201**
0.008
car_access [Yes] ref. (No)
0.0136**
0.004
bike_access [Yes] ref. (No)
0.0019
0.004
public_transport_subscription [NO] ref. (GA)
− 0.0704***
0.006
public_transport_subscription [OTHER]
− 0.0736***
0.005
public_transport_subscription [monthly ticket]
0.0384***
0.006
hh_loc_type [Rural] ref.(city_center)
0.0191*
0.007
hh_loc_type [Suburban]
0.0319***
0.007
hh_loc_type [Urban]
0.0161*
0.007
hh_income [4000 CHF or less] ref. (12,001–16,000 CHF)
− 0.0021
0.007
hh_income [4001–8000 CHF]
− 0.0087
0.005
hh_income [8001–12,000 CHF]
0.0053
0.005
hh_income [More than 16,000 CHF]
0.0130*
0.007
hh_young_kids [True] ref (False)
− 0.0185***
0.004
N (Number of observations): 22,595
LLnull (Log-Likelihood of the null model): 1481.5788999348533
LLfinal (Log-Likelihood of the fitted model): 1945.2373373249784
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
This regression analysis examines the determinants of daily travel time, revealing statistically significant associations across demographic, socioeconomic, and transportation-related variables. The results indicate pronounced gender disparities, with women exhibiting significantly lower travel durations compared to men, a finding that may reflect persistent differences in labour market participation, household responsibilities, or trip chaining behavior. Age emerges as a significant predictor, with older age groups (56–65 years and 66 + years) demonstrating increased travel durations relative to younger reference groups. This suggests that older individuals may face longer travel durations on weekdays, possibly due to slower mobility or a higher presence of discretionary trips that do not require travel time optimization, unlike commuter trips. These results shed more light on the number of weekday daily trips, indicating that individuals in the 56–65 age group take fewer but longer trips, likely due to work or commitments. In contrast, the oldest adults (66 and older) take more frequent and longer trips. This further confirms their status as Switzerland’s most active age group during weekdays. The mobility differences observed between the 56–65 and 66+ age groups can be partly explained by the transition into retirement. Adults aged 66+ are largely retired, which frees time for discretionary activities and supports higher weekday trip frequency, whereas adults aged 56–65 are often still employed and therefore concentrate their travel into fewer but longer trips. This life-course transition highlights the key role of retirement in reshaping daily mobility practices. Students may travel longer during weekdays (4% longer than employed as a reference group) because, in addition to commuting to school or university, they often engage in extracurricular activities, part-time jobs, sports, and other leisure activities.
Transportation access variables yield particularly robust effects. Vehicle-related variables (both car availability and driver’s license possession) positively correlate with travel duration, underscoring the role of private automobile use in facilitating more extensive mobility. Together with the finding that car access is associated with fewer trips, this pattern is consistent with trip chaining, where car users combine several purposes within a single journey. In the TimeUse+ dataset, trips are defined as movements between stationary activity locations (‘stays’). Under this definition, chained activities appear as fewer but longer trips, which helps explain the observed result. Bike access was not found to have significant effects. This finding complements the analysis of daily trip numbers on weekdays. Together, they suggest that people use cars less frequently but for longer trips, while bicycles are used more often but for shorter distances. Furthermore, public transport subscription data follows the same pattern as the daily trip analysis, highlighting the importance of GA or monthly tickets for commuters. Interestingly, monthly tickets had an even stronger influence than GA passes for 3.8%, suggesting that commuters rely more heavily on them and maximize their usage.
Spatial considerations reveal systematic and expected variations. Individuals residing in rural, suburban, and urban areas tend to experience longer travel times compared to those living in the city center, likely due to the greater distance to essential services and less efficient infrastructure. Suburban and rural dwellers travel 3.2% and 1.9% longer than those from the city centre. Higher-income households demonstrate elevated travel times, possibly indicative of enhanced mobility resources or greater participation in discretionary activities. Conversely, the presence of young children in households corresponds to reduced travel durations, consistent with time constraints imposed by childcare obligations. This is an interesting finding that kids do not reduce the number of daily trips on a weekday but significantly reduce daily travel time, which likely reflects compressed mobility patterns around school and caregiving schedules. Table 6 depicts the results of the GLM regression model run on weekend dataset.
Table 6
Effects of socio-economic factors on extended travel duration – weekend travel pattern
Education [Secondary education (e.g., apprenticeship or diploma)]
0.0035
0.005
Occupation [retired] ref. (employed)
− 0.0194
0.025
Occupation [student]
0.0323 *
0.015
Occupation [unemployed]
− 0.0114
0.012
driverslicense [Yes] ref. (No)
0.0030
0.014
car_access [Yes] ref. (No)
− 0.0171 *
0.007
bike_access [Yes] ref. (No)
0.0105
0.007
public_transport_subscription [NO] ref. (GA)
− 0.0144
0.010
public_transport_subscription [OTHER]
− 0.0060
0.008
public_transport_subscription [monthly ticket]
0.0262 **
0.010
hh_loc_type [Rural] ref.(city_center)
− 0.0066
0.011
hh_loc_type [Suburban]
− 0.0061
0.011
hh_loc_type [Urban]
− 0.0001
0.012
hh_income [4000 CHF or less] ref. (12,001–16,000 CHF)
0.0125
0.011
hh_income [4001–8000 CHF]
− 0.0105
0.008
hh_income [8001–12,000 CHF]
− 0.0026
0.008
hh_income [More than 16,000 CHF]
0.0429 ***
0.011
hh_young_kids [True] ref (False)
− 0.0118
0.006
N (Number of observations): 8281
LLnull (Log-Likelihood of the null model): 566.88937193471
LLfinal (Log-Likelihood of the fitted model): 631.7412123668765
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
Weekend results again indicate less presence of significant results compared to the weekday pattern. Gender has a small but significant effect, with women traveling slightly less than men on weekends. This confirmed woman being less active during the whole week. Age-related effects show that individuals aged 56–65 travel about 3% longer than the reference group (18–40), while the oldest age group (66+) does not exhibit a significant difference like in case of the weekday pattern, suggesting that older adults tend to get more rest on weekends after an active week.
Education and occupation do not show strong effects on weekend travel time, except for students, who travel 3.2% longer than other groups. This pattern likely reflects their greater schedule flexibility and higher participation in leisure activities requiring longer-distance travel. Transport accessibility also plays a role—car access significantly reduces weekend travel time, possibly because car users optimize their trips more efficiently or take shorter trips compared to those relying on public transport. Monthly public transport subscribers travel significantly more, suggesting they may use public transport for leisure trips or weekend excursions maximising monthly pass use like during the weekdays. Despite GA pass offers greater flexibility and travel options, its use is not as maximized as monthly travel pass including the whole week. Household income patterns reveal that high-income individual (16,000+ CHF) travel 4.3% longer on weekends than a reference group, likely due to greater participation in leisure activities or travel flexibility. Other tested socio-economic and spatial context factor did not prove any significant effects on weekend daily travel time. That is to say, household location showed no influence on the daily trip characteristics (both number of daily trips and travel time) during weekends.
The results suggest that weekend travel time is driven by factors beyond the tested socio-demographics and accessibility variables. While gender, income, age groups, and car access show some influence, most socio-economic factors remain unsignificant. This contrasts with weekday travel, where structured routines create clearer patterns. One key explanation is that weekend mobility is more discretionary, shaped by lifestyle choices, social activities, and external factors rather than fixed schedules. Unlike weekday commutes, weekend travel may depend more on leisure preferences, events, and even weather conditions, which were not captured in the model. The findings again align with the results for the number of daily trips, showing that weekend travel is less structured and more influenced by personal habits and external conditions, which warrants further research into qualitative aspects such as trip purposes and behavioral motivations.
Conclusion
This study leverages high-resolution GPS tracking data to shed light on the socio-economic determinants of travel behavior in Switzerland. By integrating detailed mobility records with socio-economic and spatial context factors, we explore three complementary perspectives: (1) total number of daily trips (daily trip frequency), (2) daily travel time (total time spent in transportation and (3) the choices of travel mode among those who are mobile. Our findings reveal distinct patterns not only between weekdays and weekends but also between individuals with and without car access.
The analysis of (1) and (2) underscores a pronounced disparity between weekday and weekend mobility patterns, shaped by distinct socioeconomic and behavioral factors. On weekdays, structured routines dominate: women make fewer trips due to caregiving roles, while older adults (66+) emerge as the most active cohort, balancing part-time work and leisure. Notably, households with children do not reduce trip frequency but significantly shorten travel durations, reflecting compressed mobility patterns around school runs and caregiving schedules. Similarly, individuals aged 56–65 take fewer but longer trips, likely optimizing travel around work commitments, contrasting with the 66+ group’s higher trip frequency. These findings highlight how age and caregiving responsibilities shape mobility in non-linear ways, challenging assumptions about travel behavior across life stages. Key similarities and disparities in transport mode usage further define these patterns. Cars are used less frequently but for longer trips, while bicycles enable frequent, shorter journeys—a complementary relationship emphasizing mode-specific roles in urban mobility. Public transport subscriptions reinforce this duality: monthly tickets, preferred over flexible GA passes, drive higher usage among commuters, suggesting users maximize cost-effective options for routine travel. This preference for structured ticketing aligns with weekday optimization strategies, contrasting with the decline in public transport’s weekend relevance as travel becomes discretionary.
In contrast, weekends exhibit discretionary mobility: socioeconomic factors like occupation and education fade, while leisure preferences and unmeasured lifestyle choices take precedence. For instance, high-income individuals travel more on weekends for recreation, while car owners optimize shorter leisure trips, highlighting how travel purpose reshapes behavior. Finally, the limited explanatory power of standard models for weekend travel highlights a critical research gap. While weekdays are shaped by measurable constraints, weekends appear governed by unmeasured factors like lifestyle preferences, social networks, and cultural habits. Future studies should adopt mixed-methods approaches to decode these qualitative drivers, ensuring policies reflect the full spectrum of mobility needs. By recognizing the dual nature of travel—utilitarian versus discretionary—planners can design systems that accommodate both efficiency and quality of life, fostering equitable access across all days of the week.
Nevertheless, the regression model fits and the number of significant variables in both the weekday and weekend scenarios suggest that socio-economic and spatial context factors alone do not fully explain daily mobility patterns. Clearly, other important factors, such as weather conditions (Ivanovic et al. 2023) and disruptive events e.g., public transport strikes (Yang et al. 2022) also play a significant role. Therefore, there is an emerging need to incorporate these factors into studies on mode choice and trip analysis, because this will create a more complete picture of mobility behaviors. This can be achieved most effectively with real movement data, such as GPS tracking apps, because they provide continuous, objective records of mobility that can be linked with contextual factors like weather and disruptions. While travel surveys remain valuable, their recall limitations make them less suitable for capturing the dynamic interactions between daily mobility and external conditions. In our study, we emphasize the use of GPS data, which not only captures real-world mobility behaviors more accurately, but also has greater potential to be integrated with other significant mobility factors in more comprehensive studies in the future.
In addition, our mode choice analysis reveals fundamental disparities in mode choice between car owners and non-car owners, shaped by socio-economic constraints, demographic factors, and systemic access to alternatives. Among car owners, entrenched reliance on private vehicles persists across income levels, with lower-income individuals paradoxically exhibiting stronger car dependency—a “car ownership bias” driven by fixed costs and limited alternatives. However, latent opportunities for sustainable shifts exist: women and older adults (66+) demonstrate greater openness to local public transit, cycling, and trains, particularly on weekends, while higher education and employment correlate with modest diversification into multimodal travel. These patterns suggest that car dependency is not immutable but tied to structural incentives (e.g., infrastructure gaps, habitual use) rather than preference alone.
For non-car owners, reliance on public transit and active modes is both a necessity and a constraint. Higher education aligns with public transport use over cycling, reflecting institutional access (e.g., university networks) or cost barriers to micro-mobility. Older adults again emerge as adaptive users of sustainable modes, while systemic gaps—sparse suburban/rural transit coverage, trip cost sensitivity, and childcare logistics—limit equitable mobility. Public transport subscriptions (e.g., GA passes) significantly enhance transit reliance, underscoring their role as a policy lever for this group.
Policy interventions are supposed to address these divergent realities. For car owners, reducing dependency requires more than infrastructure upgrades; incentives such as mobility budgets (subsidizing bike-sharing or rail credits) and scrappage schemes (exchanging older vehicles for transit passes) could disrupt habitual car use. Concurrently, urban design must prioritize last-mile connectivity, pedestrian zones, and safe cycling networks near residential and employment hubs. Non-car owners, meanwhile, need targeted investments in affordable, reliable public transit—particularly in underserved regions—paired with fare subsidies for low-income households. Expanding bike cargo programs and stroller-friendly transit would mitigate childcare-related barriers, while promoting flexible subscription models (e.g., tiered GA passes) could amplify existing transit engagement. Critically, temporal and geographic nuances must inform these strategies. Weekend-specific incentives (e.g., off-peak discounts) could leverage leisure-driven flexibility to promote trains and cycling, while aging-friendly infrastructure (e.g., shaded bike lanes, accessible transit stops) would sustain older adults’ high active mobility. Urban centers should prioritize bike lane expansions, suburbs require integrated rail-bus networks, and rural areas benefit from on-demand shared transport pilots.
Switzerland’s shift to sustainable transportation depends on policies that address systemic inequalities—such as unequal access to cars, income gaps, and regional disparities—while providing adaptable solutions for different groups. Decision-makers should create targeted strategies for car owners (e.g., better bike paths, incentives to use transit) and non-car owners (e.g., affordable, reliable buses and trains). By balancing weekday practicality with weekend flexibility, these efforts can build transportation systems that are fair, efficient, and inclusive for all.
Acknowledgements
The authors are grateful to swissuniversities for supporting this work through the CHORD – Swiss Open Research Data Grants (Track B, project ODTPR-SMS). They also thank the Institute for Transport Planning and Systems (IVT) at ETH Zurich for providing access to the data used in this study.
Declarations
Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Stefan S. Ivanovic
is a lecturer and a postdoc at ETH Zurich and the Center for Sustainable Future Mobility of the ETH Zurich. He holds a Ph.D. in Geographical Information Sciences from French National Institute of Geography – IGN France. His research focuses on the intersection of spatial data science and mobility behaviour, with particular interest in geospatial data quality, human mobility, and sustainable urban infrastructure.
Milos Balac
holds a Ph.D. from ETH Zurich and is currently a Senior researcher and Lecturer at the Center for Sustainable Future Mobility at ETH Zurich. His research is focused on evaluating current and future complex mobility systems and their impacts on individual mobility behavior using agent-based methodology.
Multinomial regression results for mode choice weekday trips of car owners (odd ratio)
Variable (ref. Walking)
Bike
local_pt
mpt
Train
(Intercept)
− 2.829*** (0.059)
− 6.756***
(0.001)
− 1.692*** (0.184)
− 4.961*** (0.007)
Gender Woman (ref. Man)
− 0.006
(0.994)
0.119*** (1.126)
− 0.233*** (0.792)
− 0.007
(0.993)
Age Group 41–55 (ref. 18–40)
0.074*
(1.077)
0.025
(1.025)
− 0.197*** (0.821)
− 0.377*** (0.686)
Age Group 56–65
0.289*** (1.335)
0.2***
(1.222)
− 0.183*** (0.833)
− 0.269*** (0.764)
Age Group 66 +
0.355** (1.426)
− 0.387**
(0.679)
0.149
(1.161)
1.799*** (6.043)
Higher education (ref. Mandatory education)
0.698*** (2.01)
3.776*** (43.634)
1.162*** (3.197)
0.328*** (1.388)
Secondary education
0.365*** (1.44)
3.412*** (30.33)
1.557*** (4.744)
0.084
(1.087)
Occupation—Employed (ref. Unemployed)
− 0.524*** (0.592)
1.553*** (4.728)
0.414*** (1.513)
0.376*** (1.456)
Occupation—Retired
− 0.303*
(0.739)
2.043*** (7.711)
0.091
(1.096)
− 1.768*** (0.171)
Occupation—Student
− 5.378*** (0.005)
0.739**
(2.094)
− 1.073*** (0.342)
0.069
(1.071)
Bike access—Yes (ref. No)
1.139*** (3.125)
0.174*** (1.191)
− 0.052*
(0.949)
0.299*** (1.348)
HH income 12,001–16,000 CHF(ref.under 4000)
− 0.088
(0.916)
− 0.857*** (0.424)
− 0.088*
(0.916)
0.04
(1.041)
HH income 4001–8000 CHF
− 0.2***
(0.819)
− 0.883*** (0.413)
− 0.042
(0.959)
− 0.217*** (0.805)
HH income 8001–12000 CHF
− 0.483*** (0.617)
− 1.166*** (0.312)
− 0.148*** (0.863)
− 0.249*** (0.779)
HH income More than 16 000 CHF
− 0.342*** (0.71)
− 1.352*** (0.259)
− 0.182*** (0.833)
0.012
(1.012)
HH size total 2 (ref. HH size total 1)
− 0.518*** (0.596)
− 0.111*
(0.895)
− 0.088** (0.915)
− 0.031
(0.969)
HH size total 3
0.11*
(1.117)
0.016
(1.016)
0.018
(1.018)
0.461*** (1.586)
HH size total 4
0.568*** (1.764)
0.166**
(1.18)
0.153*** (1.165)
0.658*** (1.931)
HH size total 5 or more
0.568*** (1.765)
0.293*** (1.341)
0.244*** (1.276)
0.524*** (1.689)
HH young kids-TRUE (ref. FALSE)
− 0.711*** (0.491)
− 0.282*** (0.754)
− 0.248*** (0.78)
− 0.448*** (0.639)
Public transport subscription-GA (ref. No)
0.335*** (1.398)
1.363***
(3.91)
− 1.336*** (0.263)
2.96***
(19.3)
Public transport subscription-monthly ticket
− 0.812*** (0.444)
1.603*** (4.966)
− 1.456*** (0.233)
2.045*** (7.731)
Public transport subscription-OTHER
0.261*** (1.299)
1.693*** (5.435)
− 0.135*** (0.874)
1.343*** (3.83)
Trip cost
1.086*** (2.964)
2.366*** (10.655)
1.567*** (4.795)
1.849*** (6.352)
Trip purpose-Errands (ref.Home)
− 0.621*** (0.538)
− 1.244*** (0.288)
− 0.397*** (0.672)
− 1.438*** (0.237)
Trip purpose-Leasure
− 0.487*** (0.614)
− 0.236*** (0.789)
− 0.28*** (0.756)
− 0.663*** (0.515)
Trip purpose-Other
− 0.334*** (0.716)
0.013
(1.013)
− 0.077** (0.926)
− 0.453*** (0.636)
Trip purpose-Shopping
− 0.666*** (0.514)
0.171**
(1.187)
0.355*** (1.426)
− 0.685*** (0.504)
Trip purpose-Work
− 0.215*** (0.806)
− 0.388*** (0.679)
− 0.216*** (0.806)
− 0.049
(0.953)
Trip distance
0.08*** (1.084)
− 0.079*** (0.924)
0.076*** (1.079)
0.065*** (1.067)
HH location-City center (ref.Rural)
0.573*** (1.774)
0.638*** (1.893)
0.205*** (1.228)
− 1.262*** (0.283)
HH location-Suburban
0.392*** (1.479)
0.322***
(1.38)
0.08*** (1.084)
0.42*** (1.522)
HH location-Urban
0.53*** (1.699)
0.578*** (1.782)
− 0.214*** (0.807)
0.036
(1.037)
Observations: 133,509
Ignificancy: *p < 0.05; **p < 0.01; ***p < 0.001
Log Likelihood: − 102,201.3 (df = 132)
AIC: 204,666.6
Table 8
Multinomial regression results for mode choice weekday trips non car owners (odd ratio)
Variable
Bike
local_pt
mpt
Train
(Intercept)
− 2.694*** (0.068)
− 5.909*** (0.003)
− 2.366*** (0.094)
− 6.068***
(0.001)
Gender Woman (ref. Man)
− 0.036
(0.965)
− 0.427*** (0.652)
− 0.145*** (0.865)
0.106*
(1.112)
Age Group 41–55 (ref. 18–40)
0.6***
(1.822)
0.118 (1.125)
0.318*** (1.374)
0.176** (1.192)
Age Group 56–65
0.485*** (1.624)
− 0.234**
(0.791)
− 0.331*** (0.718)
0.112
(1.118)
Age Group 66 +
1.786*** (5.965)
1.772*** (5.882)
0.037
(1.037)
1.495*** (4.46)
Higher education (ref. Mandatory education)
− 1.095*** (0.335)
0.697*** (2.008)
− 0.49**
(0.613)
− 0.089
(0.915)
Secondary education
− 0.89*** (0.411)
0.872*** (2.392)
0.227
(1.254)
− 0.229
(0.795)
Occupation—Employed (ref. Unemployed)
0.101
(1.106)
0.296**
(1.345)
0.642***
(1.9)
0.074
(1.077)
Occupation—Retired
− 2.081*** (0.125)
− 2.202*** (0.111)
− 0.698** (0.498)
− 2.146*** (0.117)
Occupation—Student
− 0.068
(0.934)
0.832*** (2.297)
− 0.133
(0.876)
− 0.004
(0.996)
Bike access—Yes (ref. No)
2.476*** (11.898)
− 0.2***
(0.819)
− 0.055
(0.946)
0.538*** (1.713)
HH income 12,001–16,000 CHF(ref.under 4000)
0.431*** (1.539)
0.366*** (1.442)
0.488*** (1.629)
− 0.254** (0.775)
HH income 4001–8000 CHF
0.204** (1.226)
− 0.007
(0.993)
0.419*** (1.521)
0.482*** (1.619)
HH income 8001–12,000 CHF
0.232*** (1.261)
0.124
(1.132)
0.599*** (1.821)
0.682*** (1.977)
HH income More than 16,000 CHF
0.627*** (1.873)
0.017
(1.018)
1.075*** (2.931)
− 0.102
(0.903)
HH size total 2 (ref. HH size total 1)
0.015
(1.015)
− 0.085
(0.919)
0.393*** (1.481)
0.367*** (1.444)
HH size total 3
0.207**
(1.23)
− 0.611*** (0.543)
0.735*** (2.086)
1.069*** (2.912)
HH size total 4
0.327*** (1.386)
0.051
(1.052)
0.415*** (1.514)
0.904*** (2.468)
HH size total 5 or more
− 0.447** (0.639)
0.512**
(1.668)
1.501*** (4.486)
0.819*** (2.269)
HH young kids-TRUE (ref. FALSE)
− 0.487*** (0.615)
− 0.468*** (0.626)
− 0.475*** (0.622)
− 0.82***
(0.44)
Public transport subscription-GA (ref. No)
− 0.221*
(0.802)
2.573***
(13.1)
− 1.417*** (0.242)
9.367*** (46.98)
Public transport subscription-monthly ticket
− 0.39*** (0.677)
3.523*** (33.878)
− 0.926*** (0.396)
9.232*** (42.15)
Public transport subscription-OTHER
0.216*
(1.241)
3.258*** (25.988)
0.174*
(1.19)
8.758*** (23.62)
Trip cost
0.285*** (1.33)
2.542*** (12.709)
1.558*** (4.748)
1.896*** (6.66)
Trip purpose-Errands (ref.Home)
− 0.969*** (0.38)
− 0.472**
(0.623)
− 0.385*
(0.681)
− 0.618*** (0.539)
Trip purpose-Leasure
− 0.673*** (0.51)
− 0.506*** (0.603)
− 0.144*
(0.866)
− 0.415*** (0.661)
Trip purpose-Other
− 0.99*** (0.372)
− 0.402*** (0.669)
0.211*** (1.235)
− 0.472*** (0.624)
Trip purpose-Shopping
− 0.534*** (0.586)
− 0.21*
(0.811)
0.221** (1.247)
− 0.133
(0.875)
Trip purpose-Work
− 0.459*** (0.632)
− 0.458*** (0.633)
− 0.201*** (0.818)
0.119*
(1.127)
Trip distance
0.054*** (1.056)
− 0.181*** (0.834)
0.033*** (1.034)
0.037*** (1.038)
HH location-City center (ref.Rural)
− 0.439*** (0.645)
0.207*
(1.23)
− 1.278*** (0.278)
0.023
(1.023)
HH location-Suburban
0.199*** (1.22)
− 0.24***
(0.787)
− 0.054
(0.948)
0.304*** (1.355)
HH location-Urban
0.321*** (1.378)
0.017
(1.017)
− 0.686*** (0.504)
− 0.583*** (0.558)
Observations: 49,554
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
Log Likelihood: − 35,557.53 (df = 132)
AIC: 71,379.06
Table 9
Multinomial regression results for mode choice weekend trips car owners (odd ratio)
Variable
Bike
local_pt
mpt
Train
(Intercept)
− 5.258***
(0.007)
− 6.094*** (0.002)
− 0.451** (0.637)
− 5.504*** (0.004)
Gender Woman (ref. Man)
0.064
(1.066)
− 0.226*** (0.798)
− 0.061*
(0.941)
− 0.02
(0.98)
Age Group 41–55 (ref. 18–40)
0.567*** (1.762)
− 0.19**
(0.827)
0.007
(1.007)
− 0.286*** (0.751)
Age Group 56–65
0.7***
(2.015)
− 0.329***
(0.72)
− 0.122** (0.885)
0.098
(1.103)
Age Group 66 +
0.225
(1.253)
− 0.275
(0.76)
0.84*** (2.316)
2.654*** (14.214)
Higher education (ref. Mandatory education)
5.322*** (24.697)
2.163*** (8.694)
− 0.123
(0.884)
0.223
(1.25)
Secondary education
5.139*** (17.617)
2.086**
(8.052)
0.037
(1.037)
− 0.106
(0.9)
Occupation—Employed (ref. Unemployed)
− 0.975*** (0.377)
− 0.457*** (0.633)
− 0.061
(0.941)
− 0.761*** (0.467)
Occupation—Retired
− 0.666*
(0.514)
− 0.213
(0.808)
− 1.107*** (0.33)
− 2.849*** (0.058)
Occupation—Student
− 5.848*** (0.003)
− 9.395***
(0)
− 1.402*** (0.246)
− 1.245*
(0.288)
Bike access—Yes (ref. No)
2.807*** (16.562)
0.41***
(1.507)
0.033
(1.034)
0.263** (1.301)
HH income 12,001–16,000 CHF(ref.under 4000)
0.13
(1.139)
− 0.082
(0.921)
− 0.085
(0.918)
− 0.783*** (0.457)
HH income 4001–8000 CHF
0.074
(1.076)
0.141
(1.152)
− 0.138*
(0.871)
− 0.72*** (0.487)
HH income 8001–12000 CHF
− 0.483*** (0.617)
0.001
(1.001)
− 0.071
(0.932)
− 1.03*** (0.357)
HH income More than 16 000 CHF
− 0.403** (0.668)
− 0.103
(0.902)
− 0.398*** (0.672)
− 0.588*** (0.556)
HH size total 2 (ref. HH size total 1)
− 0.322*** (0.725)
0.227**
(1.254)
0.191*** (1.21)
− 0.235*
(0.791)
HH size total 3
− 0.21*
(0.811)
0.045
(1.046)
0.322*** (1.379)
0.716*** (2.047)
HH size total 4
− 0.298** (0.742)
0.059
(1.06)
0.222*** (1.248)
0.238*
(1.268)
HH size total 5 or more
− 0.123
(0.884)
0.421**
(1.523)
0.521*** (1.683)
1.229*** (3.417)
HH young kids-TRUE (ref. FALSE)
− 0.217** (0.805)
− 0.236**
(0.789)
− 0.127**
(0.88)
− 0.616*** (0.54)
Public transport subscription-GA (ref. No)
− 0.541*** (0.582)
1.163*** (3.201)
− 1.198*** (0.302)
4.598*** (39.248)
Public transport subscription-monthly ticket
− 0.518*** (0.596)
1.578*** (4.845)
− 0.27*** (0.764)
3.297*** (27.042)
Public transport subscription-OTHER
− 0.008
(0.992)
1.08***
(2.944)
− 0.297*** (0.743)
2.402*** (11.049)
Trip cost
1.351*** (3.861)
2.539*** (12.667)
1.916*** (6.791)
2.355*** (10.542)
Trip purpose-Errands (ref.Home)
− 0.687** (0.503)
− 1.581*** (0.206)
− 0.532*** (0.587)
− 1.982*** (0.138)
Trip purpose-Leasure
− 0.686*** (0.504)
− 0.513*** (0.599)
− 0.502*** (0.605)
− 0.253*** (0.776)
Trip purpose-Other
− 0.651*** (0.521)
− 0.315***
(0.73)
− 0.417*** (0.659)
− 0.479*** (0.62)
Trip purpose-Shopping
− 0.888*** (0.411)
0.02
(1.02)
0.463*** (1.589)
− 0.584*** (0.558)
Trip purpose-Work
− 0.331*** (0.718)
− 0.113
(0.893)
− 0.222*** (0.801)
− 0.242*
(0.785)
Trip distance
0.061*** (1.063)
− 0.073***
(0.93)
0.05*** (1.051)
0.008*
(1.008)
HH location-City center (ref.Rural)
− 0.201
(0.818)
0.701*** (2.016)
0.154*
(1.167)
− 1.616*** (0.199)
HH location-Suburban
− 0.041
(0.959)
0.241*** (1.272)
− 0.125*** (0.882)
0.251*** (1.285)
HH location-Urban
0.393*** (1.482)
0.773*** (2.166)
− 0.118*
(0.889)
− 0.206*
(0.814)
Observations: 42,961
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
Log Likelihood: − 31,451.71 (df = 132)
AIC: 63,167.42
Table 10
Multinomial regression results for mode choice weekend trips non car owners (odd ratio)
Variable
Bike
local_pt
mpt
Train
(Intercept)
− 6.33***
(0.002)
− 2.329*** (0.097)
− 0.31
(0.734)
− 4.796*** (0.008)
Gender Woman (ref. Man)
− 0.824*** (0.439)
− 0.7***
(0.496)
− 0.152*
(0.859)
− 0.146
(0.865)
Age Group 41–55 (ref. 18–40)
0.508*** (1.662)
− 0.245*
(0.783)
0.439*** (1.55)
0.375*** (1.454)
Age Group 56–65
0.348** (1.416)
− 0.058
(0.943)
− 0.709*** (0.492)
− 0.003
(0.997)
Age Group 66 +
0.431*
(1.538)
− 2.27***
(0.103)
− 1.897*** (0.15)
− 0.666*
(0.514)
Higher education (ref. Mandatory education)
2.583* (13.233)
− 0.546
(0.579)
− 1.218*** (0.296)
− 1.533*** (0.216)
Secondary education
2.787* (16.234)
− 0.091
(0.913)
− 0.737** (0.479)
− 1.512*** (0.22)
Occupation—Employed (ref. Unemployed)
0.582*
(1.79)
− 0.509**
(0.601)
− 0.008
(0.992)
− 0.622*** (0.537)
Occupation—Retired
0.644
(1.905)
− 0.77
(0.463)
0.905*
(2.472)
− 0.488
(0.614)
Occupation—Student
0.498
(1.646)
− 0.232
(0.793)
− 0.292
(0.747)
− 0.042
(0.959)
Bike access—Yes (ref. No)
2.126*** (8.382)
− 0.455*** (0.635)
− 0.068
(0.934)
0.533*** (1.703)
HH income 12,001–16,000 CHF(ref.under 4000)
0.5***
(1.649)
− 0.928*** (0.395)
− 0.231
(0.794)
0.148
(1.159)
HH income 4001–8000 CHF
0.156
(1.169)
− 1.084*** (0.338)
− 0.527*** (0.59)
0.331** (1.393)
HH income 8001–12000 CHF
0.065
(1.067)
− 0.931*** (0.394)
− 0.316** (0.729)
0.292*
(1.338)
HH income More than 16 000 CHF
− 0.114
(0.892)
− 0.946*** (0.388)
0.477** (1.611)
− 0.519** (0.595)
HH size total 2 (ref. HH size total 1)
− 0.261**
(0.77)
− 0.585*** (0.557)
0.357*** (1.429)
− 0.34**
(0.711)
HH size total 3
− 0.133
(0.876)
− 1.59***
(0.204)
0.321** (1.379)
− 0.587*** (0.556)
HH size total 4
0.604*** (1.829)
− 0.588*** (0.556)
0.069
(1.072)
− 0.247
(0.781)
HH size total 5 or more
− 1.435*** (0.238)
− 2.144*** (0.117)
− 0.362
(0.696)
− 1.423*** (0.241)
HH young kids-TRUE (ref. FALSE)
− 0.607*** (0.545)
0.353*
(1.423)
− 0.186
(0.83)
0.029
(1.029)
Public transport subscription-GA (ref. No)
3.169*** (23.793)
2.142*** (8.518)
− 0.864*** (0.421)
4.096*** (60.121)
Public transport subscription-monthly ticket
2.286*** (9.836)
3.409*** (30.248)
− 0.793*** (0.453)
3.108*** (22.377)
Public transport subscription-OTHER
2.928*** (18.691)
3.153*** (23.398)
− 0.25
(0.779)
3.26*** (26.048)
Trip cost
1.641*** (5.159)
4.031*** (56.334)
2.955*** (19.206)
3.151*** (23.362)
Trip purpose-Errands (ref.Home)
− 0.344
(0.709)
− 0.74
(0.477)
− 0.014
(0.986)
− 2.134*** (0.118)
Trip purpose-Leasure
− 0.306*** (0.736)
− 0.189
(0.828)
− 0.048
(0.953)
− 0.102
(0.903)
Trip purpose-Other
− 0.004
(0.996)
− 0.017
(0.983)
− 0.028
(0.972)
− 0.467*** (0.627)
Trip purpose-Shopping
0.127
(1.135)
− 0.16
(0.852)
0.664*** (1.942)
0.076
(1.079)
Trip purpose-Work
0.275
(1.317)
− 1.04***
(0.353)
− 0.123
(0.884)
− 0.112
(0.894)
Trip distance
− 0.003
(0.997)
− 0.267*** (0.766)
0
(1)
0.003
(1.003)
HH location-City center (ref.Rural)
0.212
(1.237)
− 0.519**
(0.595)
− 1.074*** (0.342)
− 0.575*** (0.563)
HH location-Suburban
0.183
(1.201)
− 0.548*** (0.578)
0.329*** (1.39)
0.366*** (1.442)
HH location-Urban
0.027
(1.028)
0.013
(1.013)
− 0.335*** (0.715)
− 0.327** (0.721)
Observations: 16,327
Significancy: *p < 0.05; **p < 0.01; ***p < 0.001
Log Likelihood: − 12,360.12 (df = 132)
AIC: 24,984.24
Appendix 2: Data quality assessment
To ensure high-quality input data, we conducted an initial inspection of the dataset and identified the need for a more comprehensive data quality assessment and filtering process. As a starting point, we implemented speed filtering, which is widely recognized as a fundamental method for processing GPS mobility data (Ivanovic et al. 2019; Etienne et al. 2016).
To enhance filtering precision, separate speed filters were applied to each transportation mode, including walking, cycling, local public transport, trains, and motorized transport. Thresholds were defined based on mode-specific characteristics, with a maximum speed of 250 km/h for trains,1 160 km/h for local public transport (including S-Bahn),2 and 200 km/h for motorized transport. Despite Switzerland’s official maximum car speed limit of 120 km/h, we opted for a more flexible threshold to better reflect real-world driving conditions. This decision was further supported by the observation of a non-negligible number of car trips with recorded speeds exceeding the legal limit within the TimeUse+ data set. Another reason can be the influence of trips that started in Switzerland but continued in Germany, where some parts of the highways have no speed limit.
For walking and cycling, adjusted thresholds were proposed to account for the effects of physical exertion and fatigue. These adjustments involved reducing the upper speed limits for longer trips, reflecting findings by Dunst et al. (2018), Jones and Doust (1998), which demonstrate that speed performance decreases as travel distances increase. Specifically, thresholds were dynamically adjusted based on trip distance: shorter trips allowed for higher maximum speeds, while longer trips had progressively lower speed limits to capture the impact of fatigue. This approach ensures that filtering aligns more closely with real-world walking and cycling behaviors.
Thus, the filtering thresholds3 for walking (including running) are set as follows. For distances between:
0 and 1 km (threshold—20 km/h).
1 and 3 km (threshold—15 km/h).
3 and 10 km (threshold—10 km/h).
more than 10 km (threshold—6 km/h).
Similarly, filtering thresholds for cycling, which includes both bikes and e-bikes, are set based on the high prevalence of e-bikes in Switzerland. To account for their superior performance, the speed thresholds are adjusted as follows:
0 and 3 km (threshold—35 km/h)
3 and 6 km (threshold—30 km/h)
more than 6 km (threshold—25 km/h)
While event_id was intended to be a unique identifier, an examination of the dataset revealed instances where some event_id values appeared multiple times, with certain cases showing duplication or even the occurrence reaching 11 instances. This indicates potential data entry errors or inconsistencies in data recording. To ensure data integrity and avoid redundancy, all duplicate event_id records were identified and removed before proceeding with the analysis. About 500 different events were concerned (0.2%).
Further, we checked the attribute completeness of participants’ attributes and noticed the omission in some of them. The occupation variable contained 28 NULL values, leading to the removal of these participants from the dataset. A similar issue was found with household monthly income, where 78 participants did not provide this information. In total, we removed 100 participants, as six individuals met both exclusion criteria.
Some of the attributes also had to undergo slight modifications to provide more precise information for our models. Attribute occupation had an unspecified value ‘other’ for about 12% of participants. The "other" category in occupation lacks a clear definition and introduces ambiguity into the model, weakening interpretability. To avoid unnecessary data loss (a further 12% of the dataset) while maintaining a representative sample, we chose to reclassify rather than exclude these participants. Since individuals in this group are explicitly not classified as employed or self-employed, unemployed is the closest approximation, as they likely do not have stable or formal work. It assumes they do not participate in formal labour markets, which is likely correct. In addition, to streamline the model, we also classified self-employed participants as employed, as the primary focus of the analysis is not on distinguishing between formal employment and self-employment. While self-employed individuals may have different travel behaviors compared to traditionally employed individuals, these differences were not deemed significant for the purposes of our study, which focuses on broader socio-economic factors influencing mode choice.
Inconsistency was found in some other attributes as well. Car and bike access had four values, ‘Yes’, ‘No’, ‘No, but I can arrange to borrow one from someone (e.g., my partner, friend, neighbour)’, ‘NA’. To overcome these inconsistencies and further streamline the model, we reclassified third and fourth values as ‘NO’ because irregular access does not equate to vehicle ownership (i.e., motorbike and/or car). The ability to borrow a vehicle does not guarantee consistent access, meaning these individuals are functionally closer to those without a car. We treated ‘NA’ as uncertainty and assumed ‘NO’ as the most conservative approach. Participants who do have access to a vehicle are more likely to report it, whereas missing responses could indicate a lack of access or low relevance to their daily mobility.
Last attribute adjustment was made in case of public transport subscription. This attribute has also huge variations in values (see also Winkler et al. 2024). We reclassified values into GA,4 monthly subscription, other, and no subscription.
Benita, F.: Exploring non-mandatory travel behavior in Jakarta City: travel time, trip frequency, and socio-demographic influences. Transp. Res. Interdiscip. Perspect. 21, 100896 (2023). https://doi.org/10.1016/j.trip.2023.100896CrossRef
Bhat, C.R., Misra, R.: Discretionary activity time allocation of individuals between in-home and out-of-home and between weekdays and weekends. Transportation 26, 193–229 (1999). https://doi.org/10.1023/A:1005192230485CrossRef
Chen, C., Gong, H., Paaswell, R.: Role of the built environment on mode choice decisions: additional evidence on the impact of density. Transportation 35, 285–299 (2008). https://doi.org/10.1007/s11116-007-9153-5CrossRef
Dunst, A.K., Hesse, C., Ueberschär, O., Holmberg, H.-C.: A novel approach to the determination of time- and fatigue-dependent efficiency during maximal cycling sprints. Sports 11(2), 29 (2023). https://doi.org/10.3390/sports11020029CrossRef
Etienne, L., Devogele, T., Buchin, M., McArdle, G.: Trajectory box plot: a new pattern to summarize movements. Int. J. Geogr. Inf. Sci. 30(4), 835–853 (2016)CrossRef
Graham, S., Marvin, S.: Splintering urbanism: networked infrastructures, technological mobilities and the urban condition. Routledge (2001). https://doi.org/10.4324/9780203452202CrossRef
Harbering, M., Schlüter, J.: Determinants of transport mode choice in metropolitan areas: the case of the metropolitan area of the Valley of Mexico. J. Transp. Geogr. 87, 102766 (2020). https://doi.org/10.1016/j.jtrangeo.2020.102766CrossRef
Ivanovic, S.S., Olteanu-Raimond, A.-M., Mustière, S., Devogele, T.: A filtering-based approach for improving crowdsourced GNSS traces in a data update context. Int. J. Geo-Inf. 8(9), 380 (2019). https://doi.org/10.3390/ijgi8090380CrossRef
Ivanovic, S.S., Wood, J., Purves, R.S.: Evaluation of bicycle sharing scheme data as a proxy for cycling mobility—how COVID-19 measures influenced cycling in Paris. Transp. Res. Interdiscip.perspect. 22, 100937 (2023). https://doi.org/10.1016/j.trip.2023.100937CrossRef
Jones, A.M., Doust, J.H.: The validity of the lactate minimum test for determination of the maximal lactate steady state. Med. Sci. Sports Exerc. 30(8), 1304–1313 (1998). https://doi.org/10.1097/00005768-199808000-00020CrossRef
Kagho, G.O., Balać, M., Axhausen, K.W.: Understanding of group travel and weekend travel behavior for on-demand mobility services (Working Paper No. 1771). IVT, ETH Zurich (2022). https://doi.org/10.3929/ethz-b-000562006
Lee, Y., Circella, G., Mokhtarian, P.L.: Are millennials more multimodal? A latent-class cluster analysis with attitudes and preferences among millennial and generation X commuters in California. Transportation 47, 2505–2528 (2020). https://doi.org/10.1007/s11116-019-10026-6CrossRef
Lee, S., Ko, E., Jang, K., Kim, S.: Understanding individual-level travel behavior changes due to COVID-19: trip frequency, trip regularity, and trip distance. Cities 135, 104223 (2023). https://doi.org/10.1016/j.cities.2023.104223CrossRef
Li, W., Guan, H., Han, Y., et al.: Exploring resident’s daily activity-travel behavior: activity pattern, duration and competition. KSCE J. Civ. Eng. 25, 3121–3135 (2021a). https://doi.org/10.1007/s12205-021-2013-6CrossRef
Li, X., Wang, Y., Wu, Y., Chen, J., Zhou, J.: Modeling intercity travel mode choice with data balance changes: a comparative analysis of Bayesian logit model and artificial neural networks. J. Adv. Transp. 2021, 9219176 (2021b). https://doi.org/10.1155/2021/9219176CrossRef
Limtanakool, N., Dijst, M., Schwanen, T.: The influence of socioeconomic characteristics, land use and travel time considerations on mode choice for medium- and longer-distance trips. J. Transp. Geogr. 14(5), 327–341 (2006). https://doi.org/10.1016/j.jtrangeo.2006.01.005CrossRef
González-Sánchez, G., Maeso-González, E., López, E., Aguiar, I.: Exploring determining factors of MaaS app use and its potential effects on mobility behavior: keys to gender-sensitive planning and management. Transp. Policy 158, 175–195 (2024)CrossRef
Schlüter, J., Bossert, A., Rössy, P., Kersting, M.: Impact assessment of autonomous demand responsive transport as a link between urban and rural areas. Res. Transp. Bus. Manag. 39, 100613 (2021). https://doi.org/10.1016/j.rtbm.2021.100613CrossRef
Singh, A.C., Faghih Imani, A., Sivakumar, A., Xi, Y.L., Miller, E.J.: A joint analysis of accessibility and household trip frequencies by travel mode. Transp. Res. A Policy Pract. 181, 104007 (2024). https://doi.org/10.1016/j.tra.2024.104007CrossRef
Tukey, J.W.: Exploratory data analysis. Addison-Wesley, Reading, MA (1977)
Wang, Y., Shen, Q.: A latent class analysis to understand riders’ adoption of on-demand mobility services as a complement to transit. Transportation 51, 1043–1061 (2024). https://doi.org/10.1007/s11116-022-10351-3CrossRef
Washington, S.P., Karlaftis, M.G., Mannering, F.: Statistical and Econometric Methods for Transportation Data Analysis, 1st edn. Chapman and Hall/CRC (2003). https://doi.org/10.1201/9780203497111CrossRef
Winkler, C., Meister, A., Axhausen, K.W.: The TimeUse+ data set: 4 weeks of time use and expenditure data based on GPS tracks. Transportation 51(8), 2163–2178 (2024)
Xu, C., Wang, C.: Analysis of e-bike trip duration and frequency by Bayesian duration and zero-inflated count models. KSCE J. Civ. Eng. 23, 1806–1818 (2019). https://doi.org/10.1007/s12205-019-0674-1CrossRef
Yang, Y., Beecham, R., Heppenstall, A., Turner, A., Comber, A.: Understanding the impacts of public transit disruptions on bikeshare schemes and cycling behaviours using spatiotemporal and graph-based analysis: a case study of four London Tube strikes. J. Transp. Geogr. 98, 103255 (2022)CrossRef
