Shaping sustainable travel behaviour: Attitude, skills, and access all matter

Drawing on the conceptualisation of motility as the capacity to be mobile, this paper employs statistical and GIS-based analyses to explore the associations between travel mode choice and mobility-related attitudes, skills and opportunities to access transport modes. The study builds on survey data and spatial data from three urban contexts of Beijing, Gothenburg and Malmo to analyse both individual-level and contextual factors influencing sustainable travel behaviour. The results indicate that despite varying contexts, the three dimensions of attitude, skills and access significantly explain individuals ’ travel behaviour and their choice to travel by public transport, bicycle or car. Among the studied travel modes, cycling appears to be a competitive mode when the travel distances are within 5 km. In all three urban contexts, individuals who have greater environmental awareness are more likely to travel by public transport or cycling if the physical conditions facilitate using these modes. Good access to public transport is likely to increase the usage of both cycling and public transport and reduce car use. Favourable conditions for cycling within 2 km and 5 km radius can positively encourage people to use a bicycle as a feeder mode for public transport. Overall, our findings demonstrate that for mobility policies to increase individuals ’ motility in relation to sustainable travel modes and encourage a travel behaviour shift towards using alternatives to cars, planners need to take more holistic approaches and design policies that deal with the three motility dimensions in an integrated manner and avoid focusing on a single dimension in isolation.


Introduction
With the growing call for dealing with the climate emergency, policies promoting sustainable travel behaviour have moved to the top of the transport agendas around the world.Both researchers and policymakers have become interested in effective mobility plans that could offer a genuine alternative to cars and bring about lasting changes to individuals' travel behaviour.In this context, the current paper seeks to enrich the literature on sustainable travel behaviour and modal shift strategies by providing a comprehensive analytical perspective and a deeper understanding of the subjective factors that influence individuals' choice of sustainable travel modes.Building on spatial and non-spatial data from three urban contexts of Beijing, Gothenburg, and Mamlo, the objective of this study is to examine the effects of both contextual and individual-related factors on travel behaviour in heterogeneous urban conditions.
The study aims to bridge the fields of transport geography and mobility studies by employing the concept of motility -mobility potential -as an analytical framework and exploring both subjective and objective factors shaping sustainable travel behaviour.The paper, specifically, investigates the associations between three motility dimensions of attitudes, skills and access and individuals' mode choice in terms of using a car, public transport and cycling for daily trips.
As a sustainable travel mode, cycling is receiving more attention because of its considerable potential to replace private cars either as the main travel mode (for short trips) or in combination with public transport (to solve the 'last mile' problem of longer trips).Similarly, policies encouraging combined use of bicycle and public transport have gained attention as they can capitalise on the synergies between the two sustainable modes of transport (Bachand-Marleau et al., 2011).Integrating bicycles and public transport offers wide-ranging benefits for both individuals and communities (Buehler and Hamre, 2015).It does not only improve the quality of the environment by reducing the negative externalities of car use (e.g.greenhouse gas emissions, pollution and traffic congestion), but it also contributes to public health through increased physical activity (Krenn et al., 2015;Krizek and Stonebraker, 2011).The benefits of using sustainable travel modes are not limited to the positive health effects, as they could enhance individuals' subjective well-being and social cohesion by potentially increasing their access to life opportunities and facilitating better integration in society (Schwanen et al., 2015).However, the multiple benefits of sustainable travel behaviour cannot accrue in the absence of modal shift policies which rest on a comprehensive understanding of the complexities of changing individuals' attitudes and can alter their travel behaviour towards sustainable choices (Zhao et al., 2018).
While many mobility studies have looked into the factors affecting individuals' travel behaviour, the research on the factors that could bring about a shift to cycling or public transport or the combination of these sustainable travel modes remains insufficient.Most of the literature focuses on the effects of the built environment, urban form on travel mode choices, while the role of user-related factors such as perception, attitude, habits, and social environment is understudied (Ma and Dill, 2015;Willis et al., 2015;Zhao et al., 2018).Despite the significant methodological improvements, the transport literature often explains travel mode choice using only the assumption that a rational decision-maker automatically chooses a travel mode supported by rational motives such as the shortest travel time or the lowest costs, hence taking little notice of individuals' nuanced perceptions (Bahamonde-Birke et al., 2017).Recently, there has been a growing call for addressing this gap by incorporating subjective perspectives in transport studies and deepening our knowledge of social, psychological factors affecting mode choice and in particular sustainable travel behaviour (Kager et al., 2016).In this context, we view the concept of motility or the potential to be mobile as a helpful analytical framework in this field of endeavour.
The motility framework offers a realistic framework of modal choice that allows for incorporating a wide range of factors related to economic aspects, transport geography and social psychology in studying modal choice (De Witte et al., 2013).More specifically, it facilitates in-depth analyses of the relationships between individuals' travel behaviour and their mobility-related attitudes, norms, skills and resources.In the light of motility, an individuals' mobility is understood in relation to their characteristics and the ways they organise their daily lives, as well as contextual factors such as spatial, temporal, and social conditions influencing travel in an urban areas (Kaufmann et al., 2004;Shliselberg and Givoni, 2018).Individuals' (or groups') motility is more specifically defined using three dimensions: 1) access portfolios that they develop based on the transport options available to them, 2) their mobility-related competences, and 3) their cognitive representations and attitudes that affect appropriation of a transport mode for their daily trips (Kaufmann et al., 2004).Previous studies have attempted to identify a set of determining variables under each motility factor including, but not limited to, living location, available transport options, income and purchasing power, physical abilities, attitudes, personal and collective norms, values (e.g.environmental awareness), as well as movement-related experiences and skills (Flamm and Kaufmann, 2006;Kaufmann et al., 2018).
From the motility perspective, the key to understanding individuals' mobility choices lies in studying their personal potentials for mobility in relation to the receptiveness of the environment in which they live (Kaufmann and Doherty, 2016).Moreover, it is important to distinguish motility as the potential mobility from the realized mobility.Potential mobility comprises the entire range of mobility alternatives that individuals can act upon to carry out their daily activities, whereas realized mobility deals with the actualized part of the potential mobility -in other words, the mobility altenatives that individuals, considering the condition and circumstances, decide to utilize to complete their daily projects.In the context of this study, we focus on the part of motility and consider it as the link between motility and travel behaviour.By recognising the realized mobility as the outcome of a complex process to choose travel mode, route, destination, etc., we treat it synonymous with travel behaviour and use travel mode choice as an indicator to capture this part of motility.In employing the concepts of travel behaviour and travel mode choice we agree with the conceptualization offered by De Witte et al. ( 2013) who relate mode choice to a decision process of evaluating transport alternatives influenced by sociodemographic factors, spatial characteristics as well as socio-psychological factors.Similarly, we suggest that one can gain a more indepth understanding of an individuals' realised mobility -i.e.their travel behaviour and, in particular, mode choice by taking into account the socio-cultural and contextualised dimensions of movement.
Comparing determining factors of travel behaviour across varying socio-cultural urban contexts could shed light on how individualrelated and contextual factors shape travel behaviour and inform the policies aiming at a sustainable modal shift.To gain such insights, we study three urban areas including Beijing, China and two Swedish cases of Gothenburg and Malmo, and focus on understanding the variations across these different contexts rather than merely comparing them to each other.The reasons for our choice of study areas are twofold.Firstly, as it is often the case with research projects, our choice was influenced by the obtainable resources, availability of data and opportunities for data collection.We consider that the choice of the cases may be secondary to the main objective of the paper as long as the selected urban areas resemble different contexts.Optimally, the selected three urban areas represent a heterogeneous collection of urban contexts and allow an investigation of the effects of both individual and contextual factors on travel behaviour.Secondly, the two cases of Malmo and Gothenburg might look similar on a general level, but considering their urban scale, population Z. Hamidi and C. Zhao demographic, land use characteristics, transport infrastructure and geographic features they represent significantly different urban contexts within the same country.Especially, when it comes to the social and cultural differences, more than 46% of the population in Malmo have an immigrant background while this figure for Gothenburg is 26%.Overall, it is possible to consider the three cases in this study as examples of urban areas with three different urban scales and varying characteristics.
Beijing is a metropolitan city with 21.53 million permanent residents in 2019 within an area of 16410.54km 2 .The city has a flat topography and continental climate (Beijing Municipal Commission of Transport, 2020), which gives it a natural advantage for cycling in the city.To reach the goal of increasing the sustainable mode share -i.e.public transport, cycling and walking to 75% by 2020 since 2016, Beijing has implemented a series of intensive measures to improve conditions for public transport, walking and cycling, including building new metro lines, adding more bus lines, retrofitting 3200 km of cycling lanes by 2020 (Beijing Municipal Commission of Transport, 2016).Gothenburg is the second-largest metropolitan area in Sweden, and the municipality has over 570,000 residents (Statistics Sweden, 2019).The vision to develop a compact and carless urban area by 2035 has put an emphasis on increasing the modal share of sustainable travel modes.Particularly, despite having a hilly topography which some perceive as a barrier to cycling, the municipality aims to increase the bicycle modal share from 7% in 2011 to 21% in 2025 (Göteborgs Stad, 2019).As part of the third-largest metropolitan area in Sweden, Malmo municipality is home to about 339,000 people (Statistics Sweden, 2019).Malmo with its flat topography and oceanic climate has earned the reputation of having a bicycle-friendly environment and similar to the other two cases, has put a strong focus on promoting cycling and public transport to promote sustainable development.
In sum, the three urban areas have different cultural, social norms, economic and socio-demographic aspects, and what they have in common is a strong policy emphasis on increasing the modal share of public transport, cycling and walking for reducing the car use.The investment in increasing the modal share of public transport and cycling indicate ideal urban contexts for this study since it enriches the environmental and social environment in terms of public transport and cycling.This commonality provides reasonable grounds for incorporating the three cases in this study and investigating the effects of social structure and context on shaping individual travel behaviour.We address the objectives of the paper by investigating the following research questions: (1) How are attitudes, skills and access associated with using public transport, cycling and cars in Beijing, Gothenburg, and Malmo?(2) How does the residents' attitude in different urban contexts matter for particular mode choices such as public transport, cycling and car use?

Conceptual framework
With transport becoming a more significant domain of daily life, the theme of mobility has gained importance for understanding social and economic processes.Such momentum has led to the development of the new mobility paradigm which centres mobility at the constellations of power and highlights how having access leads to accumulation of social and economic capital (Sheller andUrry, 2006, 2016).In a similar vein, Kaufmann et al. (2004) define the potential to move in geographical (and social) space as motility or mobility capital which can be accumulated and exchanged for other forms of economic, social or cultural capital.Motility corresponds to individuals' capabilities developed by their interpretations of personal situation, needs, and wishes for mobility.Motility is decisive for how and to what degree it will actually take place and is attached to other factors such as age, gender, class etc.By incorporating individual, structural and cultural dimensions of movement in studying mobility, the motility framework allows for the examination of individuals' potential to move and their realised mobility across varying socio-cultural contexts.Flamm and Kaufmann (2006) in their qualitative study operationalised the concept of motility under three dimensions of (1) access-the range of all possible mobilities Z. Hamidi and C. Zhao according to time, place and other contextual constraints; (2) competence-the skills and abilities that directly or indirectly relate to mobilities; (3) cognitive appropriation-how individuals interpret and utilise their access possibilities and skills (Fig. 1).
The access dimension denotes the possibilities for individuals to use mobility alternatives or transport modes in the context of this study.It may strike some similarities with the concept of service as it relates to both quantitative aspects (the range of available options for travelling) and the qualities or the conditions for utilising such options (e.g.price, schedule, quality of the journey, speed, etc.).In this study, the availability of public transport stops around residents' living locations, the availability of a bike in the household as well car ownership could indicate the range of mobility options accessible to individuals to carry out their daily commutes.Comparably, the conditions that facilitate or hinder using a bicycle would reflect the qualitative aspects of the access dimension.For instance, a cycle path interrupted with several intersections between a cycle network and busy roads could indicate a higher risk of accidents and perceived barriers to cycling.Conversely, a denser and more continues cycle path network can create conditions that are more favourable for using a bicycle and encourage individuals to cycle more.
The second dimension of motility encompasses individuals' aptitudes and skills important for moving within geographical (and social) space.To carry out the act of travelling, we need both the basic skills of movement (e.g.how to walk, steer or brake) and the cognitive skills of understanding of one's static and relative positions in space (e.g.finding routes, detecting and foreseeing movement of others in space) (Flamm and Kaufmann, 2006).This emphasises the multifaceted nature of this dimension as well as its interdependencies with the other two dimensions of access and appropriation.The skills dimension of motility comprises three groups of competencies that directly or indirectly relate to utilising mobility possibilities (Kaufmann et al., 2004).The first group involves individuals' physical ability and capacities, for instance, the health conditions that allow or constrain them in cycling or driving a car.The second type is the skills that they acquire to learn and apply rules and regulations of movement, for instance, owning a driving license, knowing how to use public transport modes or even the knowledge of the different geographical areas.Finally, the third group includes the organisational skills that assist individuals in planning and synchronising activities such as how to decide on which route to take, the ability to keep up with different schedules or adapt to changes (Viry and Kaufmann, 2015).In this study, due to the data limitation, we did not examine this particular variable; however, the other two groups of variables were based on robust data.
The concept of motility through its third dimension of cognitive appropriation offers the possibility to account for the ways individuals perceive and select mobility options in relation to their own preferences and needs.The third dimension reflects how individuals interpret, evaluate and act upon both perceived or real access to travel modes and the skills they possess to use each travel mode.Such evaluation is very subjective and is shaped by individuals' needs, and plans as well as their attitudes, strategies, motives, and personal values (Kaufmann et al., 2004).In this study, we explore the cognitive appropriation dimension by looking into three factors: 1) functional suitability; 2) principle compatibility; and 3) symbolic suitability.These three factors are further elaborated in section 3.1.
It is worthwhile noting that the three dimensions of motilityaccess, skills and attitudes can evolve over time (Viry and Kaufmann, 2015) and, as Thigpen (2018) notes, this framework more explicitly acknowledges the reciprocal relationship between attitudes, norms, knowledge, skills and travel behaviour.A quantitative application of the motility conceptual framework could provide more comprehensive insights in the context of sustainable travel behaviour and advance our social understanding of mobilities (Manderscheid, 2016).

Data collection
The current study is primarily based on data collected via two survey studies in the three urban areas of Beijing, Gothenburg and Malmo.We anticipate that the differences in terms of the scale of the urban context, socio-economic, and built environment properties, as well as the transport system, could both highlight the context-specific particularities and help to generalise insights about the relationships based on the observed similarities.
The questionnaire used in the study for Beijing is same as a previous study by Zhao et al. (2018), for a detailed introduction to the survey method refer to section 3.1.of the referenced article.The data for the Swedish cases were collected using an online survey between July and October 2018.The structured questionnaire was designed based on the motility framework and the qualitative study on motility conducted by Flamm & Kaufmann (2006).After the data cleaning process, the final sample included 1075 residents of the two municipalities.51.6% of the participants were female and 48.4% male, while their age ranging from 18 to 85 averaged 48.23 (SD = 16.75).In total, 2220 respondents from the three urban areas (N = 1145 Beijing, N = 619 in Gothenburg, N = 456 in Malmo) answered similar questions regarding their attitude towards commuting by bicycle, skills and competencies related to using the three transport modes, and access to a car, public transport or a bicycle.It is important to note that in order to achieve some levels of representativeness the sample from each urban area was matched with the target population in terms of gender and age properties.
To incorporate both spatial and non-spatial aspects related to travel behaviour, we have also complemented the primary nonspatial data by spatial data retrieved from OpenStreetMap (2019)-the volunteered geographic information platform.We have, in particular, extracted and analysed the geodata on transport infrastructure including road network, public transport routes and stops.Additionally, to address the research questions, we employed statistical techniques including factor analysis, multinomial logistic regression and separately analyse three final databases corresponding to the three urban contexts.

Spatial analyses
Mobility and movement are inherently spatial.Therefore, we employed GIS-based analyses to account for the potential spatial Z. Hamidi and C. Zhao factors that could influence individuals' travel behaviour, in particular, the use of the three transport modes of interest in this study.In Beijing, the geographic centre of each neighbourhood was geocoded and assigned as the living location of the participants.In Gothenburg and Malmo, the participants living location corresponded to the centroid of their postcode arealimiting the postcode boundaries to the parts overlapping with built areas in each municipality.With regards to cycling, a set of spatial analyses were carried out to capture the bike infrastructure conditions with a comparable level of details across the three urban areas of Beijing, Gothenburg and Malmo.For this purpose, the cycle networks including both separated and shared bike paths as well as the intersections between the cycle network and major roads (including motorways, trunk roads, primary, secondary and tertiary roads) were modelled using the OpenStreet map data (2019).Next, as Figs. 2, 3, 4 present, three variables were defined to measure the ratio between the potential barriers to cycling and the facilitators of using a bicycle within three buffer radiuses of 2 km, 3 km, and 5 km around each of the geocoded locations.The three distance thresholds were set according to findings from previous studies.For example, 2 km was reported as the distance threshold for encouraging people to cycle or not in different urban contexts (Keijer and Rietveld, 2000;Zhao et al., 2018), while in one very successful cycling city, Copenhagen, the average cycling trips distance is 3.2 km, and up to 87% of all trips by bike are shorter than 5 km (DTU Data and Model Centre, 2013).The number of intersections between the cycle network and the major roads served as an indicator for accident risk, i.e. a potential barrier to cycling, while the density of the cycle network provided an indicator for bike friendliness, i.e. a facilitator for cycling offered within each studied geographical area.
In the case of public transport mode, the availability of public transport services in the participants' living environment was measured in terms of the number of public transport stations (including train stations, tram and bus stops) within 300 and 500 m radius of the geocoded locations (Figs. 5, 6, 7).

Statistical analyses
Following the spatial analysis, we employed Multinomial logistic regression analysis (MLRA) to examine the relationships between the realised travel behaviour for the three transport modes (public transport, cycling, car) and motility-related factors under three dimensions of attitude, skills, and access to the selected modes.MLRA is a statistic model that provides an indication of the relative importance of independent variables for predicting a categorical dependent variable.Since one of the outcome categories of the dependent variable serves as the reference category, the resulting coefficients indicate the odds ratio of an alternative outcome compared to the reference category (Menard, 2010).In this study, for the sake of consistency, we defined and measured both the dependent and independent variables for the three urban contexts in a similar manner; however, to address the research questions we fitted three separate MLRA models.The following sections introduce the dependent and independent variables in more details.
• 2.3.1.Dependent variables The dependent variables for the analyses derived from the participants' response to a survey question about their regularly occurring trips and the main travel mode that they use to complete such trips.The variable, in particular, captures individuals travel mode choice for trips that are most frequently made during weekdays, i.e. trips to work, school, and daily grocery shopping.Defined as a categorical variable, it records the mode choices with four nominal categories: (1) public transport, (2) cycling, (3) car and (4) other travel modes.In the MLR analysis, the last category indicating travel modes other than the three modes studied in this paper such as walking serves as the reference category.
• 2.3.2.Independent variables As Table 1 presents, the independent variables used in this study included a set of factors measuring: 1) attitude to cycling; 2) skills regarding the use of each studied mode, and; 3) access to each mode.In total, 14 same independent variables based on the three urban contexts are tested in the models.
The attitude dimension of motility was measured using three latent variables resulted from the factor analysis based on a set of questions dealing with individuals' attitudes towards cycling (Table 2, 3, 4).Factor analysis as a dimension reduction approach reduces a larger set of variables into fewer factors or components according to the interrelation of the items of shared variance (Pallant, 2013).In total, 10 similar questions capturing respondents' attitude/perception of a bicycle as a transport mode in the three urban areas were used as input for three sets of factor analyses.The respondents evaluated the 10 statements about cycling according to a 5point Likert scale (with ratings ranging from 'strongly disagree' = 1 to 'strongly agree' = 5) in Beijing and a corresponding 7-point Likert scale in Gothenburg and Malmo.As section 3.1 further details, the equivalent 10 items were reduced to three factors which are reasonably consistent across the three urban contexts and further support the choice of including the selected items (see Table 2, 3, 4).The three factors were labelled as 1) agreement on functional suitability; 2) agreement on symbolic suitability, and 3) agreement that cycling supports pro-environmental principles.These factors were then inserted in the multinomial logistic regression analysis.
While the skills (competence) dimension was measured by a set of variables describing participants' mobility skills in relation to age, gender, education, driving license and ability to cycle, the access variables captured car and bicycle ownership, access to public transport, household structure, income and cycling infrastructure.Age was considered relevant for capturing the effect of skills dimension as previous studies link age to individuals' physical ability or mental skills since older individuals might experience a decline in physical health and cognitive ability (De Witte et al., 2013;Viry and Kaufmann, 2015).Household size and income are defined as factors which reflect the access rather than skills.This is because the two factors are considered to have a stronger impact on one's access to a travel mode than the skills of using that mode.For instance, a family with one or no child may travel to a nearby destination with a bike while a family with three children may be more likely to take a bus or drive a car.Similarly, a family with one or more members who can afford to own or rent a car has a higher potential to use car compared to a family with insufficient economic resources to access a car.The two mentioned variables also reflect some aspects beyond material and geographical dimensions of access including mobility needs, plans, and organizational skills of the household members.

Factor analysisextracting components of the attitude dimension
Since the motility literature has so far dealt with the attitude dimension in a more conceptual and theoretical scope, we employed an exploratory approach to identify the underlying factors accounting for this dimension.For this purpose, we employed the 10 Z. Hamidi and C. Zhao questions introduced in the previous section, which constructed a scale measuring individuals' attitude towards cycling as a travel mode.The reliability analysis of the scale, Cronbach's Alpha coefficients, indicated an acceptable level of internal reliability for the cases (Malmo: 0.74, Gothenburg: 0.72, Beijing: 0.57).These coefficients are deemed acceptable considering the fact that the Cronbach's Alpha is sensitive to the number of items in a scale and tend to be lower for short scales with 10 or fewer items (Muijs, 2011;Streiner, 2003).Next, a set of latent factors measuring the attitude dimension were extracted using exploratory factor analysis with the principal component extraction method and the Varimax with Kaiser Normalization rotation method was applied to facilitate the interpretation of these factors.The application of the Guttman-Kaiser rule, eigenvalue larger than 1, indicated a structure with three factors accounting for more than 55% of the total variance of the data for the three urban areas.An examination of the scree plots for each case also confirmed this common three-factors structure.As Tables 2 to 4 present the retained factors (with rotated factors loadings below 0.5 suppressed), most of the items loaded strongly only on one factor except for the item on social status in the context of Gothenburg.
Because of the composition of each factor and the nature of the correlating items, we considered it fitting to refer to the three factors The first factor, in all three contexts, strongly correlates with the items capturing individuals' attitudes towards time efficiency, comfort and safety aspects of cycling.Since the mentioned items relate to the functional aspects of a bicycle and its perceived suitability as a means of travelling, we refer to this factor as functional suitability.In the context of the Swedish cases, in addition to the three primary functional aspects, two more items had strong loadings on this first factor.These items, which evaluate the possibility of using a bicycle as a means of conveyance and providing a relaxing experience during a trip, could be also interpreted as what Flamm & Kaufmann (2006) call the secondary functions of a travel mode, i.e. the role that a bicycle can take beyond self-transportation.
Three items measuring individuals' attitudes towards environmental and health impacts of cycling as well as its reliability largely contribute to the second factor.This common clustering across the three urban contexts shows that individuals evaluate this travel mode in terms of its compatibility with their environmental and health-related values and the principle reliability of travel freedom.Moreover, the observed differences in the structure of this factor across the Swedish and Chinese cases point to the influence of social collective norms on individuals attitude.More specifically, the participants in Beijing seem to consider the secondary functions of a bicycle such as the potential to carry loads or enjoying a relaxing commute beyond the typical functions of a daily travel mode, hence they are embraced as added values.
With regards to the last factor, in both cases of Beijing and Malmo the items measuring individuals' attitude towards the social image of cycling and its social status highly loaded on this factor.The cluster of these items represents the influence of cycling-related social and cultural norms as well as the extent to which an individual perceive an bicycle as a social marker.Similarly, the factor could be a measure of the suitability of bike for gaining symbolic capital; hence it was called symbolic suitability.In the case of Gothenburg, the item on the social status had slightly higher loading on the first factor, however, an association of these items with the third factor provided a more meaningful interpretation of the factors.Finally, as Tables 2 to 4 illustrate the sign and the magnitude of the factor loadings for the two items of social image and social status vary across the three urban contexts; thus affecting the nature of the third Z. Hamidi and C. Zhao factor.Considering these loadings, the factor seems to measure the symbolic suitability of bicycle in the case of Malmo, while in the case of Gothenburg and Beijing the factor designates the unsuitability of bicycle for the accumulation of symbolic capital.

Associations between motility dimensions and the use of public transport, bicycles and cars
At this stage of the analyses, three multivariate logistic regression models were fit to the data to examine the associations between motility dimensions and the use of public transport, bicycles and cars.The last category in the travel mode variable, 'Other', was set as the reference category to provide a basis for useful comparisons among the travel modes.The model fitting information presented in Table 5 and in particular the log-likelihood statistics for the final models indicate that they explain a significant amount of the variance and were good fits to the data.Diagnostic examinations and review of the Variance Inflation Factor (VIF) for the final models suggested that multicollinearity was not a significant concern (VIF < 5, Tolerance > 0.2).The analyses reveal that 14 variables measuring the three motility dimensions could be significant predictors for travel behaviour in the three study contexts.Table 5, which summarises the effects of the independent variables in the three cases, provides a basis for our discussion of the results regarding each motility dimension in the following sections.

Attitude dimension
The MLRA revealed positive coefficients between the functional suitability and using a bicycle for daily trips.This suggests that in Malmo and Gothenburg, groups who consider bicycle suitable in terms of its primary function as a transport mode are significantly more likely to use cycling over other transport modes (see Table 5 and Fig. 8).A similar pattern, although not statistically significant (P = 0.065), also seem to emerge in the context of Beijing.Conversely, respondents with a positive attitude towards the functionality of bicycle seem to be significantly less likely to use a car in all the three urban contexts.The finding that the functional suitability is associated with the increased likelihood of choosing to cycle and a lower likelihood of using car highlights the bidirectional relationship between the experience of using a transport mode and development of a more positive attitude towards that mode.Higher odds of choosing cycling as a travel mode could mean more opportunities for building cycling experiences and forming both better cycling skills and positive attitude towards this mode.In other words, more experienced cyclists are more likely to find cycling appealing and suitable to carry their daily projects.Hence, such positive perceptions of cycling functional suitability can in turn result in an increased likelihood of choosing a bicycle over other modes such as a car.In the case of Gothenburg, the residents who considered Z. Hamidi and C. Zhao cycling to be a suitable transport mode expressed a significantly lower preference for public transport as a commuting mode, which may suggest a competitive relationship between these two modes.
The results related to the second attitude factor suggest that in Beijing, those who embrace environmental and health values and find cycling compatible with principles of flexibility and travel freedom, are significantly less likely to use car compared to other travel modes.Such a travel behaviour pattern, which is also visible in Fig. 8, implies a positive association between holding such values and cycling -although the observed effect is not statistically significant.As for the Swedish cases, while in the context of Malmo holding the  mentioned values and principles does not significantly influence the likelihood of using any of the travel modes, it appears to be associated with the choice of public transport and car in Gothenburg.One possible explanation could be that public transport, similar to cycling, is often perceived as a sustainable travel mode and is, therefore, more likely to attract groups with similar values.This effect becomes more pronounced in the context of Gothenburg where, because of the hilly topography, cycling is less viewed as an optimal mode.In the case of car users, the effect may relate to the groups who agree with such principles but face other barriers (e.g.long travel distance, poor infrastructure conditions or topography too hilly for cycling), which may motivate them to use cars for their main trip.
As Table 5 illustrates, concerns for symbolic suitability of cycling seem to be a significant determinant for the likelihood of using a bicycle in Beijing.The results for the case of Malmo also indicate the same factor is positively associated with the preference for using the car over other travel modes.These effects may reflect the socio-cultural factors present in the living environment, and the influence of social environment and collective representation on individuals' attitudes towards cycling in the mentioned contexts.

Skills dimension
The results suggest that having a driving license is a significant factor for the likelihood of using public transport, bicycle or cars over other travel modes in all three contexts.In Beijing, residents who have a driving license are likely to use car or public transport more than other modes.This dual association indicates that the residents may switch from public transport to car or vice versa when they gain or lose access to a car (e.g. if they are un/able to afford a car).This trend could be associated with personal preferences or longer travel distances.Accordingly, if travelling by public transport becomes competitive with car trips, it could both reduce the incentives for public transport users to switch to using a car and encourage private car users to shift their travel behaviour.In both Gothenburg and Malmo, the residents who have a driving license are more likely to both drive and cycle.While the effect on the likelihood of using a car may be self-explanatory, its effect on cycling implies potentials for cycling and raises the prospects of sustainable travel behaviour among groups with a driving license.This implies that they may be interested in using a bicycle when the   b Intercept is also labelled as 'constant', in the regression model, it represents the average value of P when the value of all the independent variable are set as 0. c Non-significant for the presented case ernal conditions favour cycling.In other words, a current preference for commuting by car is due to barriers to cycling such as long travel distance, hilly topography, or poor cycling infrastructure.
The likelihood tests show that age is associated with the travel behaviour of the residents in Gothenburg and Malmo, but not in Beijing.In the case of Gothenburg, age is not significant for the likelihood of using public transport, cycling or cars, which indicates that it may be associated with the use of other modes (e.g.walking), and gender plays a significant role in shaping individuals' travel behaviour.Although the association is not statistically significant in any of the three modes discussed here, the coefficients presented in Table 5 suggest that the female groups are more likely to use public transport or a car and less likely to cycle for their daily commute.In Malmo, the older residents are significantly more likely to drive the car.Overall, the effect of age seems to vary in different contexts which is in line with the results from previous research on modal choice and age (De Witte et al., 2013).

Access dimension
In terms of the access dimension of motility, the indicators for access to car are significant for all three urban contexts.However, the effect of having a car is self-explanatory since it is all positively associated with commuting by car in the three cases.Similarly, as expected, the indicators for having access to a bike are associated with a higher likelihood of cycling in both Gothenburg and Malmo.
Regarding access to public transport, the study has examined access to public transport services using the density of train stations, bus and tram stops within a 300 m and 500 m radius.While access to public transport does not appear to be significant in the case of Gothenburg, higher levels of access to public transport within a 500 m radius seem to increase the likelihood of using public transport and decrease commuting by car in Beijing.Such trends raise the prospect of shifting individuals' travel behaviour towards more sustainable alternatives.A similar effect can be observed in the case of Malmo.However, the effect on cycling in Malmo suggests that the better the public transport service overlaps with a lower likelihood of cycling, which could point to a competitive relationship between public transport and cycling in the context of Malmo.
Conversely, higher levels of access to public transport services within a 300 m radius increase the likelihood of cycling in Beijing.One possible explanation for this effect is that better access to public transport (denser public transport network) often overlaps with denser urban areas, which typically have better cycling networks and a more favourable environment for cycling.Moreover, it could also indicate that in this situation, cycling may be used both as a feeder mode or a substitute mode for public transport.The size of households in Beijing appears to be significant for choosing public transport, bicycle or car over other modes such as walking.While the results suggest a positive association for all the three modes discussed here, the coefficients indicate higher likelihoods for using a car compared to a bicycle or public transport for the main trips.This effect could be due to the type and organisation of daily activities in larger households that often projects into longer or more complex daily trips and the need for accommodating joint trips.
Infrastructure conditions for cycling are found to be significant for travel behaviour in the three urban contexts.In Beijing, the poorer cycling conditions within 5 km radius, the less likely it is that individuals will choose to travel by public transport.It could indicate that poorer cycling conditions may discourage people from using a bicycle as a feeder mode for public transport, hence it reduces the opportunities for accessing public transport.This can be supported by the effect of cycling conditions within 2 km on cycling, which shows that the better cycling condition within 2 km radius increases the likelihood of travelling by bike.However, with the presence of a poor cycling condition within 5 km, people are still more likely to travel by bicycle than using other modes (which mainly includes walking), which indicate cycling is the most preferred travel mode among other modes (i.e.cars, public transport) when the travel distance is within 5 km.
The results suggest that, in Gothenburg, poor cycling conditions within a 3 km radius do not encourage people to drive either.One reason could be that in Gothenburg, public transport and walking are more preferred over cycling and driving, in other words, the trips within 3 km might be perceived too short to drive, and too hard to cycle due to the hilly topography.We can observe similar tendencies based on the modal share for short trips recorded in the recent travel survey in Gothenburg (Göteborgs Stad, 2018).
On the other hand, in the case of Malmo, the absence of optimal cycling conditions within a 3 km radius does not seem to stop people from neither cycling nor driving, meaning compared to other cases, walking seems to become less preferable when travel distances reach only 3 km.This lack of interest in walking may have to do with the individuals' perception of travel distances varying according to the scale of the urban context or the size of the urban area they use as the reference for their mental maps.The finding of an empirical study by Crompton and Brown (2006) supports that the cognitive distance of a trip in a smaller scale urban area may be perceived longer than an equal-length trip in a larger city.Moreover, the recent regional travel survey presents a consistent pattern where the modal share of walking shows sharper decreases as the travelling distances grow longer (Region Skåne, 2019).The scatter plots presented in Fig. 8 visualise the associations between the attitude factors and travel modes for the main trips, which are correspondent with the presented result above.

Discussion and conclusions
This paper applies the motility framework to explore how three dimensions of mobility-related attitudes, skills and access to transport modes are associated with individuals' travel behaviour in Beijing, Gothenburg and Malmo.The motility framework, by incorporating both individuals' characteristics and contextual conditions in analysis of travel behaviour, appears to have made statistically significant and conceptually substantive contributions to our understanding of individuals' modal choice.Overall, our findings support that motility dimensions, in three varying urban contexts, can significantly explain individuals' travel behaviour, and thus play important roles in shaping sustainable travel behaviour.
The operationalisation of the concept of motility revealed that three factors underlying individuals' attitudes towards cycling as a daily travel mode include: perceived functional suitability, principles compatibility and symbolic suitability.Our findings revealed that the three attitude factors matter in the three urban contexts to different extents, and the aspects influence each attitudinal factor may vary across different urban contexts.
In general, among the three factors, functional suitability seems to play an important role for the participants from all three urban contexts.While all three factors are associated with travel behaviour in Beijing, symbolic suitability is not associated with the case of Gothenburg, and principle compatibility does not impact the travel behaviour in Malmo.Perceived functional suitability has influenced the choice of all the three modes in Gothenburg, while in Malmo it impacts the choice of both cycling and car use, and in Beijing, it is only associated with cycling.Principles compatibility only impacts the car mode choice in Beijing but it is associated with both public transport and car using in Gothenburg.When it comes to the symbolic suitability, it does affect the cycling mode choice in Beijing and car driving behaviour in Malmo, but has no influence to mode choices in Gothenburg.
To summarize, both the aspects influencing the three attitude factors and the effect of the factors show variations across the cases.With respect to the perceived functional suitability of cycling, in Beijing, speed, safety and comfort were considered as functional requirements to evaluate a bicycle as a suitable travel mode, while in the Swedish cases, the requirements further extended to the secondary functions such as the potential to carry loads or enjoy a relaxing experience.The composition of the factor denoting symbolic suitability of cycling highlight the differences between the Swedish and the Chinese cases in terms of the influences of their social context or collective norms.Our results suggest that in Beijing, the association of cycling with low social status raises concerns over the social image that using a bicycle may convey, meaning the collective social norms seem to discourage choosing cycling as a daily travel mode.Conversely, in Gothenburg and Malmo, the opposite signs of the factor loading for the two items of social image and social status of cycling imply an inverse relationship between their effects, suggesting that the Swedish social environment has a less discouraging effect on individuals' decision to cycle.Overall, our findings from the factor analyses indicate that the two Swedish cases share certain similarities in terms of individuals' attitudes towards cycling and the collective representation of bicycle as a daily travel mode.More specifically, there is a closer resemblance between the factors underlying the attitude dimension of motility in Malmo and Gothenburg compared to the case of Bejing.
Our results demonstrate that attitude factors can significantly explain individuals' travel behaviour and are consistent with the finding of previous studies (Lind et al., 2015;Friman et al., 2019) which report significant associations between individuals' attitude, personal belief or social norms and their mode choice.In particular, perceived functional suitability of bicycle seems to be a significant factor behind increased likelihood of choosing cycling over other travel modes in Swedish cases and a lower likelihood of car use in all three urban contexts.In general, utilisation of a travel mode seems to depend on individuals' perception of whether or not that travel mode and its functions could satisfy their needs.As such, modal shift policies success could lie in both improving the (primary and secondary) functions of sustainable travel modes and creating awareness and a positive perception of such functions.To improve the primary and secondary functions of sustainable travel modes, the target groups should be carefully considered, they could be noncyclists who are often car users, or casual cyclists who would like to use bike as a means of conveyance or to have calm and stressfree commutes.Municipalities or companies can initiate behaviour nudging through pilot projects to provide an opportunity for the residents or employees to try the improved functional sustainable travel modes and create awareness by sharing promotional information as well as interacting with the residents who are involved in the pilot projects.For example, it is a broadly introduced approach in Scandinavian countries that cities or companies procure bicycles/e-bikes and offer them to the residents or staff -who usually travel with car -for a trial period of three months or half year.Other initiatives include subsidy schemes on electric bikes, provision of different types of bikes and ensuring suitable parking spaces for them to meet the different needs and uses of bikes, e.g.travelling with kids, pet or carrying loads.Prominent examples include offering cargo bikes to families with kids or free bike rides for the elderly to provide them with the possibility of visiting recreational destinations by bike instead of a car.Such trials are often planned with involving researchers who can follow participants' behaviour change and gain deeper insights through interactions with the participants.Meanwhile, publicizing such pilot projects via social media and creating a positive image of cycling would also raise awareness and contribute to the normalization of bike as travel mode, hence affecting the social environment.The attitude factor of compatibility with principles and values also play a significant role for individuals' choice of travelling by a sustainable travel mode.In all three contexts, residents with greater environmental awareness or health values, if not faced with barriers such as poor infrastructure, seem to prefer using a bicycle or public transport.This accords with the findings of a study by Abrahamse et al., (2009) indicating that stronger personal norms could lead to stronger behavioural intentions, however, the intentions can turn into actions if (real or perceived) barriers do not exist.Nevertheless, raising public awareness of environmental and health values combined with removing barriers to using sustainable travel modes could be instrumental for realizing intentions and shaping sustainable behaviour.
The results on the effects of mobility-related skills on individuals' travel behaviour are in line with previous findings (Cass and Faulconbridge, 2016) that mastering the competencies of using certain travel modes is associated with shifting to these travel mode.Our analyses reveal that individuals, who have driving skills, besides travelling by car are also likely to take up cycling.Although driving skills are mainly used for travelling by car, exercising such skills requires a certain level of mental and physical abilities, hence having a driving license may also provide an indication of individuals' capacities for developing cycling skills.From a policy perspective, our results suggest that the group with a driving license seem to be multimodal users who are likely to travel by car, public transport or bicycle.This implies that obtaining a driving license and taking the option to drive does not necessarily mean that people prefer to travel by car exclusively.While lack of access to a car may be one of the reasons behind choosing to travel by other modes, individuals may decide to do so because of practical advantageous of those modes.Therefore, sustainable mobility policies could capitalise on this multimodal behaviour and encourage this group to become regular users of sustainable travel modes.Previous studies (Hernandez and Monzon, 2016;Steg, 2003;Wang et al., 2013) advocate that measures aiming to make public transport services competitive with cars -e.g. by improving reliability, frequency, physical and spatial accessibility, or designing transport hubs as attractive public spaces -combined with policies seeking to alter the cultural and psychological values associated with car.Reducing the efficiency/dependency of driving not only could retain the current public transport users but even more to encourage car users to start using public transport.It is worth noting that, as Redman et al.,(2013) find in their study of attractive public transport qualities, most attributes that could effectively encourage a modal shift are context-specific and affective, thus connected to individual perceptions, motivations and should be reviewed from case to case.In the case of cycling, studies by Strömberg and Karlsson (2016) and Thigpen (2018) provide evidence that short-term bicycle use also offers the potential to bring about lasting positive changes in individuals' travel behaviour -particularly by improving individuals' cycling-related perceptions, attitudes and skills.
Other key results highlight that the extent and the level of access to the studied transport modes have a significant effect on individuals' travel behaviour in all three urban contexts.Overall, cycling is a competitive mode among others when the travel distance is within 5 km, however, better cycling conditions within 5 km radius seem to encourage people to use a bicycle as a feeder mode for public transport.In large scale urban areas such as Beijing, public transport and cycling may compete with each other for providing the possibility for people to travel with inter-modality.Responding to Kuhnimhof et al., (2010) 's study, they suggested that the radius for cycling around residential locations is between 3 and 5 km, which however depends on the urban contexts.This suggests that sustainable modes may become more preferable to cars when the travel distance is less than 5 km in megacities such as Beijing.In regards to optimum travelling distances promoting intermodality, to integrate cycling and public transport, the study indicates that the distance from origins to the public transport should not be further than 2 km, and the optimal cycling condition is essential for encouraging people to travel with cycling as a feeder to connect to public transport.Future urban planning could coordinate both land use and transport policies dedicated to reducing commuting distances and improving both accessibility and quality of public transport.Our findings support that good access to public transport not only could reduce car use by attracting public transport users but it has positive impacts on cycling, meaning a promising opportunity of developing multimodal in the form of a combination of cycling and public transport.One reason for the increased likelihood of using a bicycle in the presence of better access to public transport could be that policies aiming to improve access to public transport may overlap with the policies designed to decrease the convenience of using a car.In general, reducing the reliability of car together with raising public awareness about sustainable alternatives to car and removing the perceived barriers to using such alternatives could lead to the formation of multimodal travelling behaviour (Bamberg et al., 2003).While embracing a positive attitude towards sustainable modes may encourage using them, the actual shift towards Z. Hamidi and C. Zhao sustainable travel behaviour can only happen if other barriers, e.g.long travel distance, poor infrastructure for cycling, limited access to public transport are removed (Abrahamse et al., 2009).As it has been addressed by previous studies (Cao and Hickman, 2019;Cuthill et al., 2019;Song et al., 2019), the overall effect of good access to the travel modes emphasises the importance of providing accessible sustainable mode services for encouraging sustainable travel behaviour.
In conclusion, having a positive attitude to sustainable travel modes and perceiving them suitable for carrying out daily trips is significant to individuals' decision to use them as daily travel mode.However, such factors alone in the absence of skills or a receptive and favourable environment may not bring about lasting behavioural changes.Thus, for mobility policies to successfully increase individuals' motility and potentials for using sustainable travel modes, planners need to take more wholistic approaches and design policies that deal with the three motility dimensions of attitudes, skills and access in an integrated manner and avoid focusing on a single dimension in isolation.This study further confirms the potentials that the motility framework offers for building a comprehensive understanding of people's travel behaviour, however, more research is needed to fully realise such potentials in the context of planning practice.Future research could focus on the further examination of the links between individuals' potential and the receptiveness of their environment through operationalisation of this concept at varying spatial levels, i.e. considering micro, meso and macro-level contextual conditions.Other avenues for future research could include identifying travel mode-specific factors under the attitude dimension and exploring the attitude factors important for intermodality.

Fig. 3 .
Fig. 3. Bicycle infrastructure in Gothenburg in relation to study areas.

Fig. 4 .
Fig. 4. Bicycle infrastructure in Malmo in relation to study areas.

Fig. 5 .
Fig. 5. Public transport infrastructure in Beijing in relation to study areas.

Fig. 6 .
Fig. 6.Public transport infrastructure in Gothenburg more comprehensive insights study areas.

Fig. 7 .
Fig. 7. Public transport infrastructure in Malmo in relation to study areas.
Likelihood: 1798.68**− 2 Log Likelihood: 1110.24**− 2 Log Likelihood:883.90**Note: the reference category is walking; * P < 0.05; **P < 0.001.a B: value used for calculating the probability of a case falling into a specific category.It indicates the direction of the relationshipwhich factors increase (when the value is positive) the likelihood of a 'yes' answer and which factors decrease it (when the value is negative).

Fig. 8 .
Fig. 8. Associations between the attitude factors and travel modes for the main trips in three urban contexts.

Table 1
Description of all the tested variables in the MLRA model.

Table 2
Components and scores of attitudinal questionnaire items for Beijing.
Z.Hamidi and C. Zhao

Table 3
Components and scores of attitudinal questionnaire items for Gothenburg.Varimax with Kaiser Normalization.Rotation converged in 5 iterations.Kaiser-Meyer-Olkin Measure of Sampling Adequacy = 0.86, N = 619, % of total variance explained = 66.17.

Table 4
Components and scores of attitudinal questionnaire items for Malmo.

Table 5
Parameter estimates of the effect of attitude, skills and access variables on travel modes for main trips in Beijing, Gothenburg and Malmo.