The effects of life-events and changes in mobility tool ownership on mode choice behaviour

The effects of life-events and changes in mobility tool ownership on mode choice behaviour | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article The effects of life-events and changes in mobility tool ownership on mode choice behaviour Roel Faber, Sander van Cranenburgh, Maarten Kroesen, Eric Molin This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5353959/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Mode choice is an essential subject within travel behaviour research. Typically, mode choice has been studied using cross-sectional (discrete choice) models, which assume that all choices are made simultaneously. In this study, we relax this assumption by explicitly incorporating the time when a choice is made within the modeling framework, using a latent transition choice model. This model allows for the identification of the effects of life-events and (changes in) mobility tool ownership on mode choice probabilities over time. To estimate the model, data from the Netherlands Mobility Panel gathered between 2016 and 2022 are used. The model identifies two latent classes, 1) a car-dependent modality style and 2) a multi-modal modality style. The transition probabilities between these classes in-between two consecutive waves are estimated, as well as the effects of life-events and mobility tool ownership on these transition probabilities. We find substantial and statistically significant effects from changes in vehicle ownership on the transition probabilities, indicating that electric bicycle ownership leads to a substitution of the car towards the bicycle on shorter-distance trips even after controlling for lead- and self-selection effects. Behavioural change Discrete Choice Modelling Modality Styles Mobility Biographies Netherlands Mobility Panel Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Mode choice analysis is a fundamental subject within travel behaviour research (McNally, 2000 ). The dominant approach to modelling mode choice behaviour has been based on utility-maximization discrete choice theory since the 1970s (McFadden, 1974 ; Train, 2009 ). Traditionally, discrete choice models are employed in a static fashion, meaning that they do not consider changes in preferences over time (Ben-Akiva et al., 1997 ). Given the habitual nature of travel behaviour (Aarts & Dijksterhuis, 2000 ; Gärling & Axhausen, 2003 ) this assumption seems reasonable. Habits can often be difficult to break, frustrating policymakers’ efforts to change travel behaviour. But behavioural changes do occasionally occur (Schwanen et al., 2012 ; Strömberg et al., 2016 ). Moreover, from a policy perspective, understanding when individuals change their travel behaviour and travel preferences is vital information for crafting policies to achieve desired behavioural changes. Such periods of change can be seen as ‘windows of opportunity’, where policies can have a larger impact on behavioural change. One research stream focusing on behavioural change is the mobility biographies framework, where the study of behavioural change has focused on life-events, such as residential relocations, starting families, or changing jobs (Müggenburg et al., 2015 ; Rau & Manton, 2016 ). In this literature, these life-events are then seen as windows of opportunity for behavioural change (Scheiner, 2017 ). Previous work has indeed shown that people’s habits are indeed prone to change during these events (de Haas et al., 2018 ; Müggenburg et al., 2015 ). Aside from life-events, which are more general changes as one goes through life, research has also investigated the effects of mobility tool ownership, such as vehicle or public transport card ownership, on travel behaviour (Loder & Axhausen, 2018 ; Nurul Habib et al., 2018 ; Scott & Axhausen, 2006 ). However, these studies have only looked at the effect of life-events on mode use (Gao et al., 2023 ), rather than on mode choice. Mode use here is defined as the result of both the trip generation process (which trips will a certain person make) and the mode choice process (given a certain trip, which mode will a certain person choose). Since mode use combines both trip generation and mode choice, studies that focus on mode use, for example in terms of total kilometres or number of trips travelled with certain travel modes, cannot disentangle possible changes in trip-generation from changes in mode choice preferences. Furthermore, these studies typically use clustering approaches(de Haas et al., 2018 ; Olde Kalter et al., 2021 ) as opposed to discrete choice modelling techniques. Consequently, they have not been able to study how attribute tastes and preferences change as a result of life-events. In a separate research stream emerging in the last decade, some transport researchers have developed longitudinal choice models to study mode choice behaviour (Xiong et al., 2015 ; Zarwi et al., 2017 ). These longitudinal choice models enable the study of evolving preferences over time. However, this previous work however did not focus on the effects of multiple changes within the individual, such as life-events. Zarwi et al. ( 2017 ) studied the effects of changes to the transportation system, whereas Xiong et al. ( 2015 ) looked at one set of general ‘life-stages’. This paper intends to further combine the fields of mobility biographies, including mobility tool-ownership, and the longitudinal study of mode choice behaviour. The main aim is to study how individual characteristics, including life-events and changes in mobility tool ownership, shape mode choice preferences over time. To achieve this objective, we will employ the relatively rare latent transition choice model. This model is an extension of a normal latent class choice model, where time dynamics are explicitly considered in the class-membership function and individual respondents’ transitions between classes over time are modelled explicitly. As a result, it is a relatively parsimonious way to effectively model the change of mode choice preferences over time. The model utilizes panel data from the Netherlands Mobility Panel (MPN), which is enriched with the life-events and changes in mobility tool ownership. 2. Literature overview and Conceptual Model In this section, we will introduce the relevant literature and use it to build a conceptual model that will guide the analyses in this paper. The literature overview consists of three building blocks: first, we will provide an overview of studies into behavioural change. Second, we will look at mode choice analysis and the idea of modality styles and third, we provide an overview of the literature on longitudinal choice models in travel behaviour research. Finally, we synthesize the findings from these building blocks in a conceptual model. 2.1. Behavioural change in transportation One subject area within travel behaviour research is that of the inertia that is present when people make decisions. This area works under the assumption that decision makers’ choices are driven to some extent by habits, contrasting the usual assumption of full rationality for each new choice situation that underlies typical choice models (Aarts & Dijksterhuis, 2000 ; Gärling & Axhausen, 2003 ; Neal et al., 2012 ). The formation and breakdown of these habits are key topics of interest. Moments when habits are broken down are seen as the key windows of opportunity for changing people’s travel behaviour. Two such potential ‘habit-breaking’ moments are life-events and changes in mobility tool ownership (Clark et al., 2014 ; Gao et al., 2023 ; Janke & Handy, 2019 ). Life-events are key events in one’s life course that entail a disruption to day-to-day life (de Haas et al., 2018 ; Müggenburg et al., 2015 ), therefore breaking habitual travel behaviour. Such events can be related to family life (childbirth, marriage, leaving the home, residential relocations) and employment (gaining or losing employment, gaining or losing working from home abilities; Gao et al., 2023 ). These life-events can prompt a person to re-evaluate their habitual behaviour and thus provide a critical window of opportunity to enact behavioural change (Janke & Handy, 2019 ; Kløckner, 2013 ). Up to this point, the longitudinal studies on life-events in travel behaviour research has focused on cluster analyses (de Haas et al., 2018 ; Kroesen, 2014 ). These analyses reveal shifts in travel patterns, but they are unable to show changes in attribute preferences and elasticities as a result of the life-event. Another area of travel behaviour research focuses on so-called mobility tools: vehicles, drivers’ licenses, and (discount) passes for public transport (Scott & Axhausen, 2006 ). The acquisition of these tools is based on longer-run expectations of mobility needs. Simultaneously, these tools allow their owners to more easily, cheaply, and/or effectively use certain transport modes. As a result, they create lock-in effects (Scott & Axhausen, 2006 ): owning a car makes the car a more attractive option even for trips that might be more suitable to the bicycle or public transport. Consequently, mobility tool ownership is closely linked to the existence of travel habits and could be a crucial explanation for the modality styles introduced above. 2.2. Mode Choice analysis and modality styles Discrete choice modeling has been a cornerstone of travel behaviour research since the introduction of random utility-maximization theory within discrete choice analysis (McFadden, 1974 ). Two impactful applications of these models are the estimation of the value of travel time, typically using stated preference data based on route choices (Small, 2012 ; Wardman et al., 2016 ), and modeling of mode choice in large-scale travel demand models (Ben-Akiva & Lerman, 1985 ; McNally, 2000 ; Train, 2009 ). One key improvement on the discrete choice theory workhorse, the multinomial logit model, is that of the nested logit model (Carrasco et al., 2002 ; Daly & Zachary, 1979 ; Williams, 1977 ). The nested logit model groups subsets of alternatives which are similar in some unobserved characteristics, which enables it to remove the irrelevance of independent alternatives (IIA) property of the multinomial logit model. This improvement is highly relevant in the case of mode choice analysis, as some modes compete more with other similar modes than they do others (Train, 2009 ). Owing to the exponential increase in computing resources, the investigation of preference-heterogeneity within the population rose to the forefront of modeling efforts since the 1990s. For example, much interest has been paid to the distribution of the value of travel time across the population (Cirillo & Axhausen, 2006 ; Fosgerau, 2006 ; Hensher & Greene, 2003 ; van Cranenburgh & Kouwenhoven, 2021 ). Simultaneously, for mode choice analysis, research has shown how preferences for various modes vary across the population (Bhat, 2000 ; Cherchi et al., 2017 ). One concept that helps to communicate heterogeneity in mode choice preferences is that of modality styles. Modality styles are discrete segmentations, based on variations between people in underlying preferences to use certain modes (Diana & Mokhtarian, 2009 ; Molin et al., 2016 ; Vij et al., 2013 ). Typically, these modality styles are uncovered using clustering methods, where people are grouped based on the number of times they make use of certain travel modes (Faber et al., 2022 ). Another method to identify these modality styles is the latent class choice model (Faber et al., 2022 ; Keskisaari et al., 2017 ; Prato et al., 2017 ; Vij et al., 2013 ). Using this method has the advantage that it ties the concept of modality styles into discrete choice theory. 2.3. Longitudinal Choice Modeling and Latent Transition Choice Model Traditionally, choice models have been employed in a static, time-indifferent, fashion (Train, 2009 ). These models are agnostic to the process or order of the choices made and therefore operates under the implicit assumption that (from the modellers’ perspective) all choices are made more or less simultaneously. As a result, the time when a choice is made is not considered in the modelling process. We want to note that for many, perhaps even the vast majority of choice modelling, this assumption is completely valid (Ben-Akiva et al., 1997 ). However, this static paradigm does entail that the analyst is unable to determine the effects of changes in characteristics, either on the level of the trip or the decision-maker, on the choice probabilities over time (Hamaker, 2012 ). Instead, differences between respondents can be used to estimate the effects of certain characteristics. These effects are then often assumed to be similar to longitudinal changes within respondents over time, for example when the choice model is used to forecast future travel demand after some changes have been made. The problem with this method is that it is uncertain whether the between- respondent effects are similar to the effects within respondents. Take for example the potential effect of electric bicycle ownership. Within a static approach, a model might estimate the effect of e-bike ownership on bicycle choice probability by comparing the difference in choice probabilities across respondents that either do or do not own such a vehicle. However, it is not unlikely that respondents who own an electric bicycle are more avid cyclists in the first place, which prompted them to buy an e-bike. There is thus likely to be a rather strong self-selection effect, making it difficult to ascertain the effect that electric bicycle ownership would have on the choice behaviour of people who do not yet own such a vehicle. Two potential approaches to solve this problem are commonly found in the literature: first, and most prevalent, is the use of stated preference data. These analyses explicitly ask respondents to make choices considering hypothetical scenarios, where the researcher is free to design the experiment, and they can vary the attributes or scenarios associated with certain choices. In our example, respondents could be prompted to make choices between the bicycle and the car, first without considering electric bicycle ownership and then in the hypothetical scenario where they did own such a vehicle. However, as stated preference choices are made ‘on paper’ they suffer from a potential lack of external validity (Louviere et al., 2000 ; Murphy et al., 2005 ). For example, respondents might not be familiar enough with an electronic bicycle to give reliable estimates of how owning such a vehicle would change their mode choice behaviour. The second solution is to use multiple measurements per individual, ideally taken across a wide enough range of time, and to then explicitly model the time when the choices are made within the modelling context (Ben-Akiva et al., 1997 ). One example of such a choice model is the latent transition choice model, also known as the Markov choice model (Ben-Akiva et al., 1997 ; Liao et al., 2018 ; Xiong et al., 2015 ; Zarwi et al., 2017 ). The latter name derives from the Markov process, where the probability of each event only depends on the state of the previous event and not on any previous states. The former name, which we prefer to use in this article as it aligns more closely with common nomenclature in the field of travel behaviour research, originates from latent transition cluster modeling, which has been a popular approach to study longitudinal data in the field of travel behaviour research for some time (see for example Kroesen, 2014 ). The model works by allowing the class membership probabilities, estimated on the level of the individual, to change over time (Anderson, 1954 ; Böckenholt & Langeheine, 1996 ; Wiggins, 1955 ). These changes, or transitions, can then be affected by external factors or other changes within the individual, such as life-events or changes in mobility tool ownership. As a result, these models allow for the modeller to estimate the effects of changing circumstances or explanators of choice behaviour on the level of the individual. It therefore allows us to estimate the effect of certain characteristics on choice behaviour within individuals. 2.4. Conceptual Model The relationships in the literature mentioned above are graphically summarized below in Fig. 1 . This conceptual model is then used to further guide the analyses in the paper. To build the conceptual model, we start with the discrete choice building block, which assumes that observed mode choices can be explained using both the observed and unobserved attributes of each travel mode alternative for the trip. In this study, we will use alternative specific travel times and travel distances as these attributes. We then assume that distinct modality styles exist, and we let the effect of the attributes on mode choice vary across these modality styles. Then we allow socio-demographic factors, life-events, and mobility tool ownership to affect the class membership probability of these modality styles. Following the reasoning that the modality styles are largely inert, class membership in the first wave affects class membership in the second wave. Finally, we include a moderating effect of life-events, (changes in) mobility tool ownership, and socio-demographic factors on the probability that someone switches between modality styles. 3. Research methods 3.1. Mathematical Model The latent transition choice model builds on the standard latent class choice model, with the addition of latent transition parameters, which are used to estimate the class transition probabilities between the longitudinal waves. For an overview of the mathematical definition of the latent class choice model, the reader is referred to Ben-Akiva et al. ( 1997 ) and Hess & Daly ( 2014 ). For the latent transition choice model, the log-likelihood function for observing a series of choices k made by decision maker n belonging to class s at timepoint t , with alternatives i , can be written as a function of taste parameters \(\:{\beta\:}_{s}\) . The taste parameters are thus conditional on decision maker n belonging to class s at wave t in Eq. 1: $$\:LL\left(\beta\:\right)=\sum\:_{n\:=\:1}^{N}ln\prod\:_{t\:=\:1}^{T}\sum\:_{s\:=1}^{S}{\pi\:}_{n,t,s}\left(\prod\:_{k\:=\:1}^{K}{P}_{n,t}\left({i}_{k,t}|{\beta\:}_{s}\right)\right)$$ ( 1 ) The key addition here is that the class-membership probability π is dependent on the wave t at which the choice was made. The class-membership function of the latent classes for wave 1 follows conventional standards and is estimated as a multinomial logit function based on an initial class-specific constant \(\:{\delta\:}_{s}\) , as well as a function \(\:g\) of a vector of parameters \(\:{y}_{s}\) and a vector of socio-demographic characteristics, life-events, and (changes in) mobility tool ownership \(\:{z}_{n,t}\) , as given in Eq. 2. $$\:{\pi\:}_{n,s,t\:=\:1}\:=\:\frac{exp({\delta\:}_{s}\:+\:g({y}_{s},\:{z}_{n,t}\left)\right)}{\sum\:_{s}exp({\delta\:}_{s}\:+\:g({y}_{s},\:{z}_{n,t}\left)\right)}$$ ( 2 ) The class-membership function of the latent classes for the second wave, however, is specified to be conditional on the class-membership probability of the first wave. The transitions between the classes are then modelled as in Eq. 3, where the probability that a decision maker n who belonged to class r in wave 1 will belong to class s at wave 2 is equal to the transition probability \(\:{tr}_{n,s,r}\) of class r to class s , multiplied by the class-membership probability of belonging to class r in wave 1: $$\:{\pi\:}_{n,s,t\:=\:2}\:=\sum\:_{s=1}^{S}{\sum\:}_{r=1}^{R}{\:(tr}_{n,s,r})\:\left({\pi\:}_{n,t\:=\:1,\:\:r}\right)\:\:\:$$ ( 3 ) These transition probabilities themselves are modelled as multinomial logit functions as well, such that the transition probability depends on a transition parameter \(\:{{\phi\:}}_{s,r}\) associated with the transition from class r at the previous wave to class s at the current wave and a function \(\:g\) of both parameters \(\:{y}_{s,r}\) and a vector of sociodemographic characteristics, life-events and (changes in) mobility tool ownership \(\:{z}_{n,t}\) , as given below in Eq. 4: $$\:{tr}_{n,s,r}=\:\frac{exp({{\phi\:}}_{s,r}\:+\:g({y}_{s,r},\:{z}_{n,t}\left)\right)}{\sum\:_{s}exp({{\phi\:}}_{s,r}\:+\:g({y}_{s,r},\:{z}_{n,t}\left)\right)}$$ ( 4 ) 3.2. Research Data We use trip data from the travel diary of the Netherlands Mobility Panel ([MPN], for more information see Hoogendoorn-Lanser et al., 2015 ), a household panel in the Netherlands that comprises an extensive questionnaire and a 3-day travel survey. Respondents for the MPN are recruited from the Kantar NIPObase, an invite-only internet access panel (IAP). Invitations for the Kantar NIPObase are sent out based on register data. Members of the larger NIPObase IAP are then invited for the MPN separately, based on their socio-demographic characteristics. Between 30 and 50% of respondents from the larger IAP decide to join the MPN upon receiving an invitation. When respondents have entered the MPN, their yearly response rates for each wave vary around 85%. We use data from the yearly waves between 2016 and 2022. For each unique respondent, we select one set of two consecutive waves. If there were sets of consecutive waves where life-events or changes in mobility tool ownership happened between the two waves, then we always selected one of these sets. This procedure ensures that the final dataset contains as many life-events and changes in mobility tool ownership as possible. If no life-events or changes in mobility tool ownership happened, then one set of consecutive waves is drawn at random for each person. The sample descriptives for the final sample, as collected during the first wave used in the dataset, is given in Table 1 . Table 1 Sample descriptives compared to population distribution Sample (%) Population (12 + inhabitants of the Netherlands, 2019; %) Gender Male 48 50 Female 52 50 Age (Years) 12–24 14 18 25–44 30 28 45–64 35 33 65 + 21 21 Education Low 30 34 Medium 36 40 High 34 26 Urban Density residential municipality (addresses/m 2 ) 2500 22 25 Household Type Single 22 20 Only adults 31 46 Adults and children 46 34 The sample descriptives are very similar to the population values for nearly all variables. The only exception is household type, as our sample consists of comparatively more households with children. This is most likely the result of the biased sampling procedure introduced above, where we purposefully oversampled sets of waves that include life-events. As life-events are more common within households with children, these types of households will be oversampled as well. This enables us to identify the effects of life-events on the transition probabilities more reliably. The primary unit of analysis of the mode choice model is the trip. Principally, we analyse trips as recorded by respondents in each waves’ travel diary. However, not all trips in the travel diary are useful for our analysis. Therefore, some selection criteria are used. First, only trips that departed from the residential location were selected, as the residence is typically the location where the mode choice decision is made. Second, all trips made with modes other than the car, public transport, the bicycle, or on foot were discarded. The shares of the other modes are marginal, and estimating valid attribute-parameters for them is therefore not feasible. Third, trips for which a very large distance (> 200km) was reported are excluded, as the decision-making process for such trips differs from that of the more typical, daily trips. These selection criteria leave us with a total number of 28,117 trips made by 4,789 unique respondents. 3.3. Operationalisation We use alternative-specific travel times and travel distances as the trip-specific explanatory variables in the model. These travel times and -distances are calculated using the Google Directions API based on the origin, destination, and departure time of the trip. In the utility function, we use both a linear and a square root transformation of travel time for each mode to capture possible non-linear effects of travel time on the utility of each mode. Travel distance is used in the utility function of the active modes. Finally, aside from alternative specific constants for all trips we also use a dummy-variable for trips that are made with other people. This dummy variable is used to correct for the fact that such shared trips are more often made with the car compared to the other modes, allowing for better estimates of the travel time and travel distance parameters. This dummy variable is kept fixed across the latent classes. The final utility functions then are given in Eq. 5 below. Note that the parameter pertaining to travel distance is only estimated for the active modes. $$\:Utility=\:asc+dumm{y}_{shared\:trip}+\:{\beta\:}_{travel\:time}+\:{\beta\:}_{\sqrt{travel\:time}}+{\beta\:}_{travel\:distance}\:$$ ( 5 ) The utility functions are kept relatively simple, to balance with the complexity of estimating the transitions between latent classes. Estimating both a very complex utility function and the transition parameters would quickly lead to an over-specified model. For the travel times and travel distances, which are the key explanatory variables in the utility functions, descriptive statistics are given in Table 2 . Table 2 Descriptive statistics for the alternative-specific travel times and travel distances of all trips in the dataset Min. Mode Mean. Max. Travel Time (min.) Car 1 7 12.5 141 Public Transport 1 22 34.5 2234 Bicycle 1 10 31.9 671 Walking 1 33 114 2373 Travel Distance (km) Bicycle 0.1 2.9 10 212 Walking 0.1 2.6 9.2 195 In the class-membership and transition functions, we use socio-demographic characteristics, life-events, and (changes in) mobility tool ownership. For the life-events and (changes in) mobility tool ownership, we have used the variables given in Table 3 , which are presented together with their absolute and relative occurrence in the sample. The correlations between these variables are not very high, with the exception of the variables pertaining to car ownership and access to cars. The highest correlation here however is 0.68, and therefore there are no strong concerns regarding multicollinearity. Table 3 List of life-events and mobility tool ownership variables and their occurrence in the data Occurrence (N; %) Life-events Residential relocation 290 (6.1%) Change of job 558 (12%) Birth of a child 147 (3.1%) Shift to working from home 204 (4.5%) Mobility Tool Ownership Car ownership 3,405 (71%) Always has access to car 2,910 (61%) Never has access to car 915 (19%) E-bike ownership 955 (20%) Owns Personal Public Transport Card 3,282 (69%) Changes in Mobility Tool Ownership Gained personal car 260 (5.5%) Lost personal car 223 (4.6%) Gained access to car 303 (6.3%) Lost access to car 220 (4.6%) Gained electric bicycle 515 (11%) Gained public transport card 181 (3.8%) Lost public transport card 175 (3.7%) 4. Results This section presents and discusses the main results. First, we show the goodness-of-fit of increasingly complex models, starting with a multinomial logit model and ending with a latent transition choice model which includes covariates. We show that the latter model provides the best fit to the data. Then, the class-specific parameters of this model are introduced, and we show that the latent classes can be interpreted as modality styles, as they reflect underlying predispositions to use certain travel modes. We will then discuss the class-membership results, which is followed by a discussion of the transitions between the two classes and the substantive meaning of these transitions. 4.1. Model Selection The choice models are primarily estimated in Apollo (Hess et al., 2019 ; Hess & Palma, 2019 ). As the model is non-trivial to implement in Apollo, we validated the results by using a separate implementation in Matlab. To test whether the latent transition choice model offers an empirical benefit over more parsimonious models, we estimated six models, each increasing in complexity. Table 4 contains the goodness-of-fit statistics of these six models: 1) A MNL model 2) A nested logit model 3) A conventional latent class choice model 4) A latent class choice model that allows for the sizes of the classes to vary per wave 5) A latent transition choice model 6) A latent transition choice model with covariates From the second model onwards, a nesting structure is used, with one nesting level. The root level contains the car and one nest, which contains the alternatives public transport, the bicycle, and walking. This nesting structure is based on the idea that public transport, the bicycle, and walking substitute each other more directly than they do the car. A comparison of the second model with the first model reveals whether this nesting structure improves model fit. Based on the literature and the conceptual model, we expect this to be the case for mode choice analysis on our revealed preference dataset. A more complicated nesting structure was tested where the active modes were separated into a further subnest, but this structure offered no improvement. The third model explores whether people’s preferences for mode alternatives are heterogeneous by estimating two separate latent classes. For the sake of parsimony in what is otherwise already a relatively complex model, we decided to fix the number of classes to two for all latent class models. Adding additional latent classes would result in an exponential growth of possible transitions and, therefore, transition parameters. Given the relatively modest probability of transitions between classes, we think that the dataset is not large enough to support more than two latent classes. The fourth model can capture behavioural changes across the population, which would result in different class sizes for the two waves. However, this model is not able to estimate which individuals’ behaviour has changed. The fifth model, the latent transition choice model, is an improvement in that respect, as it can now assess which individuals transition between the estimated latent classes. Then, finally, we report a model (model 6) with covariates of both the initial class-membership and transition probabilities. This model enables an estimation of factors that influence the transition probabilities. These allow for much richer behavioural interpretations of the results. Table 4 Goodness-of-fit statistics for the choice models Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 MNL-model Nested Logit model 2 class latent class model 2 class latent class model. change size across waves 2 class latent transition choice model. No covariates 2 class latent transition choice model. With covariates Est. parameters 10 11 20 21 22 80 Goodness-of-fit statistics N (individuals) 4 789 N (choices) 28 117 LL 0 -37 923 LL β -20 421 -19 941 -17 350 -17 350 -17 279 -16 682 Mean LL β per person -0.724 -0.706 -0.617 -0.617 -0.616 -0.586 ρ 2 eq. shares 0.462 0.474 0.543 0.543 0.544 0.560 ρ 2 obs. shares 0.314 0.331 0.418 0.418 0.420 0.440 AIC 40 863 39 904 34 740 34 742 33 937 33 525 BIC 40 946 39 994 34 905 34 915 34 784 34 185 Cross-validation LL β per obs. In sample -0.731 -0.715 -0.622 -0.622 -0.619 -0.598 LL β per obs. Out of sample -0.706 -0.689 -0.600 -0.600 -0.597 -0.578 % Diff. -3.57% -3.61% -3.52% -3.52% -3.53% -3.37% As can be seen in Table 4 , the nested logit model (model 2) provides a much better fit to the data than the MNL model 1, indicating that there are nesting structures in the mode choice data. Furthermore, the nesting parameter lambda, is found to be significantly different from 1 (not reported in Table 4 ), which also indicates the presence of nesting structures. The latent class models, starting with model 3, outperform the standard models, indicating the presence of mode choice heterogeneity. Model 4 does not provide a better fit to the data than model 3, meaning there are no substantial changes in mode choice behaviour across the population between the two waves. The transition model, model 5, however, provides a statistically significant better fit than both previous latent class models (LRT = 142, df = 1, p < 0.001). It also performs better in the 5-fold cross-validation tests, where the dataset is split into five parts and the model is subsequently estimated using four parts and tested on the remaining hold-out part. The model that uses covariates to determine the class-membership and transition probabilities (model 6) then statistically outperforms the latent class transition model without covariates (model 5) as well (LRT = 1193, df = 58, p < 0.001). This model will be used for further examination in the remainder of this paper. 4.2. Interpreting the latent classes as modality styles Table 5 reports the class-specific parameter estimates. Note that these parameter estimates, which are conditional on class-membership, are stable over time and thus the same for both waves. Table 5 Estimated class-specific model parameters Class 1 Class 2 Car PT Bicycle Walking Car PT Bicycle Walking ASC - -1.77 (-9.15) 0.76 (5.16) 3.83 (18.7) - 1.22 (9.65) 3.67 (22.7) 5.23 (13.1) Shared Trip dummy - -0.98 (-10.4) -0.89 (-13.0) -0.322 (-4.39) Same as class 1 Travel time 0.0176 (9.04) 0.026 (7.27) Square root travel time -0.910 (-10.6) -0.86 (-10.05) Square root travel distance - -0.389 (-10.8) -0.996 (-11.1) - -0.220 (-13.1) -0.712 (-6.50) Nesting parameter 0.725 (-7.45 a ) 0.493 (-12.0 a ) In this and following tables, robust t-ratios given between parentheses a: Robust T-ratio with respect to 1 We find that the alternative-specific constants (ASCs) are very different for the two classes, which supports the notion that both classes represent modality styles that reflect underlying predispositions towards the use of certain travel modes. The ASCs of public transport, bicycle, and walking are all smaller for the first than the second class. This provides a first indication that the first class is in general more inclined towards the use of the car than the second class. Both travel time parameters are positive, but both square root travel time parameters are negative. The net effect of these parameters is that all modes’ elasticities with respect to travel time are negative for both classes, as is to be expected. Similarly, travel distance has a negative effect on active mode use. To provide a more intuitive picture of the mode choice differences between the two classes, we have calculated the conditional probabilities for each of the two classes for trips falling within several ‘distance’ bands. To do so, we used the average travel time per mode for all trips within the dataset that corresponded to a given distance. Note that this procedure introduces some noise, particularly for the longer distances, which average out travel times and travel distances over fewer trips. As a result, there are some fluctuations in the graph for distances lager than 10km. The estimated conditional probabilities for both classes and distances between 0 and 20 km are given in Fig. 2 . Figure 2 shows that there is a large difference in conditional choice probabilities between the two classes. For the first class, the probability of using the car increases rapidly as distances increase. The car becomes the dominant mode around 2.5km and then the choice probability asymptotically increases towards a choice probability nearing 1 for trips longer than roughly 10 km. For the second class however, the bicycle is the dominant mode up until trips of roughly 10 km in length. At that point the car becomes the most probable mode, followed by public transport. Based on these conditional mode choice probabilities for each class, we can identify the first class as ‘Car-oriented’ and the second class as ‘Multimodal’. In the estimated choice model, roughly 60% of people belong to the ‘car-oriented’ class and 40% to the ‘multimodal’ class in the first wave. 4.3. Class-membership function Now that the latent classes have been identified as two distinct modality styles (‘car-oriented’ and ‘multimodal’), we interpret the class-membership function. The parameters estimated in the class-membership function show the relation between class-membership and socio-demographics, life-events, and (changes in) mobility tool ownership. These parameters are given in Table 6 . Importantly, the class-membership function is specific to membership in the first wave, before any transitions might have happened. The life-events and changes in mobility tool ownership have thus not happened yet. The direction of the causal effect therefore is not clear: either people with certain behaviours are more likely to undergo life-events/changes in mobility tool ownership or people might already be aware of upcoming life-events and have already changed behaviour accordingly. Thus, there might be both selection effects (first option) or lead-effects (second option). Table 6 Parameter estimates of the class-membership function for the first wave Class 2: ‘Multi-modal’ (ref: class 1, ‘Car- oriented’) delta -1.06 (-2.36) Socio-demographics Age < = 24 (ref: age 65+) 1.11 (5.13) Age 25–44 (ref. age 65+) -0.0293 (-0.163) Age 45–64 -0.129 (-0.841) Employed 0.0828 (0.0653) Works from home -0.05 (-0.439) Low income (ref: med income) 0.269 (2.64) High income (ref: med income) 0.172 (1.15) Children in household 0.0790 (0.65) Urban Density -0.0547 (-0.84) Life-events (between wave 1 and wave 2) New Job -0.124 (-0.86) Started working from home 0.0793 (0.363) Child Born -0.564 (-2.20) Residential Relocation 0.063 (0.33) Mobility Tool Ownership , wave 1 Personal Car -1.12 (-6.02) Access to Car, always -1.91 (-13.9) Access to Car, never 0.443 (1.40) Drivers’ License -2.01 (-5.50) Electronic Bicycle 0.938 (7.84) PT-card 0.396 (3.69) Changes in Mobility Tool Ownership (between wave 1 and wave 2) Lost personal car 0.363 (1.84) Gained personal car -0.676 (-2.80) Lost access to car a 1.33 (6.95) Gained access to car a -0.892 (-4.06) Gained electric bicycle 0.840 (5.88) Gained PT-card 0.395 (1.85) Lost PT-card -0.42 (-1.51) The socio-demographic and mobility tool ownership parameters show that the class-membership estimates are congruent with the earlier identification of the classes as modality styles. People who own a car, a driver’s license, and always have access to a car are more likely to belong to the car-oriented modality style. Conversely, people who own electronic bicycles or public transportation cards are more likely to belong to the multi-modal modality style. Similarly, as expected, young people and people with lower incomes are more likely to belong to the multi-modal modality style as well. Furthermore, the class-membership estimates reveal some interesting effects of life-events and changes in mobility tool ownership on modality style membership. Prospective parents were more likely to belong to the ‘car-oriented’ modality style before the child was born, even after controlling for the effects of age. Similar lead- or selection effects can be found for most changes in mobility tool ownership. People who bought a personal car between waves 1 and 2 were already more likely to belong to the ‘car-oriented’ modality style than would otherwise be expected, as were people who gained the ability to always access a car. A similar but opposite effect is found for the addition of an electric bicycle: people who gained an electric bicycle in between the two waves were already more likely to belong to the multi-modal modality style. 4.4. Transitions between modality styles The main advantage of the latent transition choice model is that we can combine the above within-class mode choice probabilities with calculations regarding the transition probabilities between the latent classes. Furthermore, the model calculates to which extent the transition probabilities between the classes are affected by life-events and changes in mobility tool ownerships. Below, we first present the statistically significant transition parameters and the socio-demographic, life-event, and mobility tool ownership parameters in Table 7 . For the sake of parsimony, variables without any significant effect are not shown in the table. Following a short discussion of the existence and direction of some effects below the table, we will illustrate the effects of a selection of life events and changes in mobility tool ownership on transition probabilities. Table 7 Parameter estimates of the transitions (non-significant parameters not shown) Transition ‘Car-oriented’ to ‘Multi-modal’ Transition ‘Multi-modal’ to ‘Car-oriented’ Transition parameter ( \(\:{\phi\:}\) ) -7.84 (-3.71) -5.50 (-4.53) Socio-demographics Age < = 24 (ref: age 25+) 1.50 (2.80) n.s. Life-events New Job 0.785 a (1.86) n.s. Started Working from Home 0.66 (2.05) n.s. Higher Income -1.91 (2.06) n.s. Residential Relocation n.s. 0.84 a (1.77) Mobility Tool Ownership in first wave Access to Car, always -1.08 (-2.38) n.s. Access to Car, never n.s. -1.32 (-1.97) Electronic Bicycle 1.26 (3.44) n.s. PT-card 1.71 (2.84) n.s. Changes in Mobility Tool Ownership Gained personal car n.s. 2.51 (2.05) Gained access to car n.s. 2.84 (2.63) Gained electric bicycle 1.76 (4.35) n.s. a: only statistically significant if we accept a 10% threshold for significance testing Both transition parameters are negative and statistically significant, indicating that transitions between classes are relatively rare events. There is a very limited effect of static socio-demographic variables on the transition probabilities. This is to be expected, given that the effects of static socio-demographic covariates on class membership are already estimated in the class-membership model of the first wave. There are some effects of life-events, although they are not very strong. Interestingly, both working from home and starting a new job are associated with shifts away from the car-oriented modality style, although both effects are not very strong. This however does indicate that people start making different mode choices after they start working from home, which is a relevant finding given the large increase in working from home during and after the COVID-19 pandemic. There are relatively strong effects of both mobility tool ownership in wave 1 and changes in mobility tool ownership between the two waves on the transition probabilities. When interpreting these coefficients however, we must be careful regarding the assumed direction of causality. For example, we might interpret the negative coefficient (-1.08) of ‘always having access to car’ in wave 1 on the transition from car-oriented to multi-modal modality styles in two distinct ways: first, that always having access to a car prevents people from making this transition. This assumes that mobility tool ownership is a cause of our travel behaviour and our changes therein. Second, that people who are otherwise disinclined to make such a transition, and thus probably are relatively car-dependent, are more likely to ensure they always have access to a car. This assumes that our travel behaviour, and especially our habitual patterns, causally affect our mobility tool ownership. In practice, both directions are likely to exist (Nurul Habib et al., 2018 ; Scott & Axhausen, 2006 ). Below, we will try to keep both options in mind, but for reasons of legibility will not discuss both options for each coefficient. The first thing to notice is that all effects are in the expected direction, given the interpretation of the latent classes as modality styles: people who always have access to a car are less likely to transition from the car-oriented to the multi-modal modality style, whereas people who own electric bicycles or public transport cards are more likely to do so. Similarly, the effects of changes in mobility tool ownership also follow the expected direction. These effects are relatively strong as well: both gaining ownership of a personal car (2.51) and the closely related gaining the ability to always access a car (2.84) make it much more likely that someone transitions from the multi-modal to the car-oriented modality style. Gaining ownership of an electric bicycle makes people more likely to switch from car-oriented to multi-modal modality styles (1.76). These transition parameters can be used to calculate transition probabilities. Using these transition probabilities, we can create transition matrices, which are shown in Table 8 . Table 8 Transition matrices for respondents without life-events and for those with changes in mobility tool ownership People without life-events or changes in mobility tool ownership People who gained a car Wave 2 Wave 2 Class 1 Class 2 Class 1 Class 2 Wave 1 Class 1: Car- oriented 0.939 0.0609 Class 1: Car-oriented 0.967 0.0337 Class 2: Multi-modal 0.0723 0.928 Class 2: Multi-modal 0.264 0.736 People who lost a car People who gained an electric bicycle Wave 2 Wave 2 Class 1 Class 2 Class 1 Class 2 Wave 1 Class 1: Car- oriented 0.847 0.153 Class 1: Car-oriented 0.833 0.167 Class 2: Multi-modal 0.0337 0.967 Class 2: Multi-modal 0.00929 0.991 As can be seen in the upper-left quadrant of Table 8 , the class membership of people who do not undergo any life-events or changes in mobility tool ownership is very stable, as roughly 94% and 92% of car-oriented and multi-modal people remain in their respective modality styles. This picture shifts dramatically for those who gain or lose a car and for those who gain an electric bicycle. Two things stand out: first, that gaining a car leads to a relatively higher chance to transition towards a car-oriented modality style (26%) than losing one leads to a transition towards a multi-modal modality style (15%). Car-oriented behaviour therefore seems to be more stable than multi-modal travel behaviour, and there is some asymmetry in the effect of vehicle ownership. Second, that gaining an electric bicycle is fairly effective at getting people to transition towards a multi-modal modality style. A final result with respect to the transition probabilities is that we find relatively weak effects of life-events on transition probabilities. Perhaps these life-events don’t lead to changes in mode choice , but only to changes in mode use . This can be the result of changes in the activity-pattern generation, for example by affecting either trip generation or trip distribution rather than mode choice. Residential relocations to urbanized areas for example might lead individuals to make more shorter-distance trips, which are more likely to be made using the bicycle and walking. However, there need not be a change in sensitivity to travel distance and thus no large change in behavioural parameters. Given the same trip, the respondent would still make roughly the same choices. However, the types of trips made might have changed. Another explanation might be that life-events coincide with changes in vehicle-ownership. As we explicitly model the effects of vehicle-ownership, this indirect effect will not show up in the model. However, the corelations between life-events and changes in vehicle ownership were relatively small (< 0.2). Therefore this explanation is unlikely to fully explain the weak effects of life-events. 4.5. Enumeration of transition effects on choice probabilities To get a more intuitive understanding of the meaning of the transitions between the modality styles, we used the estimated LC transition model to calculate mode choice probabilities for individuals with varying states of vehicle ownership. As a result, we can see the effect that changes in vehicle ownership have on the mode choice probabilities. We illustrate this result using two different approaches. First, to show the effect of changes in car access, we show the estimated choice probabilities for trips made by people grouped by car ownership. As people’s car ownership changes, so does their estimated probability of belonging to a certain modality style. Due to these transitions between the modality styles, their estimated choice probabilities change as well as is shown in Fig. 3 . A couple of observations can be made based on Fig. 3 . First, car ownership is a large determinant of travel behaviour: people who own a car are much more likely to choose the car. Second, the effect of buying a car is much larger than the effect of selling a car. The increase in car probability is much larger for people who bought a car than the decrease for people who sold a car. Third, there are substantial lead effects: people who buy a car between the two waves already use the car much more often than people who do not do so. For the second illustration, we again use the reference trips within each kilometre band, which we used earlier to illustrate the difference between the modality styles. Now, we plot the unconditional bicycle choice probability of people grouped by e-bike ownership. This allows us to show how the probability of choosing the bicycle across various distance ranges changes as a result of differences in e-bike ownership. We want to highlight two results here. First, that buying an electric bicycle increases the probability of choosing the bicycle. For the two groups whose electric bicycle ownership did not change, we observe no change in mode choice probability for the bicycle. For the group that did buy an e-bike, we observed a sizeable increase in the predicted market share of the bicycle from the first to the second wave. Model estimates of the treatment effect of the untreated, that is, people who did not already buy an electric bike, of buying an electric bicycle are roughly in the range of 8 percentage points, with the mode share of the bicycle increasing from 33–41%. This increase corresponds with a decrease in the estimated mode share of the car. As such, our results indicate that the electric bicycle substitutes car use, especially for shorter-distance trips up to roughly 15km in length. 5. Conclusion In this paper, we used a latent transition choice model to estimate the longitudinal transitions between modality styles. Two latent classes are found, which can be identified as two distinct modality styles, namely ‘Car-oriented’ and ‘Multimodal’. The ‘car-oriented’ modality style is found to be more sensitive to travel time increases than the multimodal modality style. The modality styles are relatively stable over time, especially in the absence of any life-events or changes in modality styles. The car-oriented modality style is found to be more stable than the multi-modal modality style. Life-events only have relatively minor effects on the transition probabilities between the modality styles: people who started a new job or increased the hours they worked from home moved from the car-oriented to the multimodal class slightly more often. These findings seem to contradict earlier studies showing relatively larger effects of life-events on travel behaviour (Clark et al., 2014 ; Gao et al., 2023 ; Olde Kalter et al., 2021 ; Rau & Manton, 2016 ). These previous studies typically used clustering approaches, which cannot disentangle trip generation changes from mode choice changes. Our results therefore suggest that life events mostly affect trip generation rather than the mode choice process itself. Ownership of mobility tools and changes therein have much larger effects on the transition probabilities. Whilst the existence and direction of these effects are not wholly surprising (car ownership increases the probability one belongs to the car-oriented modality style, and electric bicycle ownership increases the probability one belongs to the multi-modal modality style), some findings are still very noteworthy. First, we find clear evidence for an asymmetry in the effect of car ownership, where gaining a car has a much larger effect on the transition probabilities than losing a car. As a result, attaining ownership of a car seems to have an irreversible effect on one’s travel behaviour, even if the car has to be sold. From a policy perspective – assuming that reducing overall car use is the policy objective – it is therefore important to facilitate lifestyles that do not depend on car ownership and provide car-less people with similar levels of accessibility as car owners. Second, we find evidence of either lead- or self-selection effects, where people who buy a car or bicycle in-between the two waves already respectively used the car or the bicycle more often in the first wave than those who did not buy such a vehicle. These findings point to the importance of establishing good counterfactuals when studying the effects of vehicle ownership on travel behaviour, for example by using longitudinal data. Third, we find that, even after controlling for these selection effects, buying an electric bicycle results in a noticeable shift towards the more multi-modal modality style. The transition towards the less time- and distance sensitive multi-modal modality style also suggests that buying an electric bicycle makes bicyclists less sensitive to increases in travel time and travel distance. As a result, the bicycle choice probability increases substantially, especially for shorter-distance trips up to roughly 15km in length. These findings complement earlier studies using longitudinal clustering and structural equation modeling methods (de Haas et al., 2022 ; Kroesen, 2017 ). The latent transition choice model thus enables us to improve further our understanding of the effects of changes in mobility tool ownership on mode choice behaviour. However, even though the model captures dynamic effects and thus uses the possibilities of panel data, the causal direction is still difficult to establish. This is due to the yearly occurrence of the data, which means that changes in mobility tool ownership and changes in choice behaviour seem to coincide together. This ambiguity makes it more difficult to draw clear behavioural conclusions. A second drawback of the current approach relates to the relatively low occurrence of life-events and changes in mobility tool ownership. As a result, the power of the model to reliably assess the effects of these changes on mode choice behaviour is limited. The estimation of the model is also made more difficult, as it is dependent on relatively limited observations where changes in choice behaviour occur. A final limitation we would like to highlight is that the current model is only estimable with two latent classes. As the number of latent classes increases, the number of transitions between classes increases exponentially. To illustrate, even a three-class solution would require the estimation of six transition parameters. Combined with the difficulties mentioned above, this was not feasible. The downside of a two-class solution is that they might oversimplify the existing heterogeneity with respect to mode choice behaviour. There are several areas for future research that seem worthwhile. First, we could estimate a model using more longitudinal waves. This could prove especially fruitful given the relatively low occurrence of life-events. The ability to use more data could enable us to provide more reliable estimates of the effects of changes in life-events on travel behaviour. Second, we could add more detailed life-events and perhaps model changes of the transport system as well. Examples of the first type could include whether a residential relocation moved towards a more car-oriented or a more multi-modal oriented residential environment and a related example of the second type is to study whether autonomous changes in the built environment have an impact on the transition probabilities. More generally, longitudinal choice models can be used to estimate the stability of preferences regarding attributes such as time and cost, which could be interesting to empirically estimate longitudinal effects of, for example, changes in income on the value of travel time. Finally, use of more intensive longitudinal data could allow for a better understanding of the directions of causality involved. The yearly waves in the MPN are unable to capture when exactly behavioural shifts occurred, and if they preceded or followed life-events or changes in mobility tool ownership. Daily longitudinal data, for example using GPS tracking devices, combined with more complex travel surveys could help to further our understanding of the causality involved. Declarations Funding The authors received no specific funding for this work Author Contribution RF - Conceptualization, Methodology, Software, Formal Analysis, Data Curation, Writing - original draft, VisualizationSvC - Methodology, Software, Writing - Review & Editing, SupervisionMK - Conceptualization, Writing - Review & Editing, SupervisionEM - Conceptualization, Writing - Review & Editing, Supervision Data Availability Access to data from the Netherlands Mobility Panel can be requested through https://mpndata.nl/. The specific data file used to support the analyses in this paper, as well as the scripts necessary to reproduce it from the base data files in the Netherlands Mobility Panel, are saved by the corresponding author. Competing Interests The authors have no conflicts of interest, financial or otherwise, to declare that are relevant to the content of this article. References Aarts, H., Dijksterhuis, A.: The automatic activation of goal-directed behaviour: the case of travel habit. J. Environ. Psychol. 20 (1), 75–82 (2000). https://doi.org/10.1006/JEVP.1999.0156 Anderson, T.W.: Probability Models for Analyzing Time Changes In Attitudes. Free Press. (1954). https://scholar.google.nl/scholar?hl=nl&as_sdt=0%2C5&q=Probability+Models+for+Analyzing+Time+Changes+In+Attitudes&btnG = Ben-Akiva, M., Lerman, S.R.: Discrete choice analysis: theory and application to travel demand, 1st edn. MIT Press (1985). https://trid.trb.org/view/274564 Ben-Akiva, M., Mcfadden, D., Abe, M., Böckenholt, U., Bolduc, D., Revelt, D., Steinberg, D.: Modeling Methods for Discrete Choice Analysis. Mark. Lett. 8 , 273–286 (1997) Bhat, C.R.: Incorporating Observed and Unobserved Heterogeneity in Urban Work Travel Mode Choice Modeling. Https://Doi.Org /10.1287/Trsc.34.2.228 . 12306 , 34 (2), 228–238. (2000). https://doi.org/10.1287/TRSC.34.2.228.12306 Böckenholt, U., Langeheine, R.: Latent change in recurrent choice data. Psychometrika. 61 (2), 285–301 (1996). https://doi.org/10.1007/BF02294340/METRICS Carrasco, J.A., De Dios Ortúzar, J., De, J., Ortu, D., Zar, Â.: Review and assessment of the nested logit model. Transp. Reviews. 22 (2), 197–218 (2002). https://doi.org/10.1080/01441640110091224 Cherchi, E., Cirillo, C., de Ortúzar, J. D: Modelling correlation patterns in mode choice models estimated on multiday travel data. Transp. Res. Part. A: Policy Pract. 96 , 146–153 (2017). https://doi.org/10.1016/J.TRA.2016.11.021 Cirillo, C., Axhausen, K.W.: Evidence on the distribution of values of travel time savings from a six-week diary. Transp. Res. Part. A: Policy Pract. 40 (5), 444–457 (2006). https://doi.org/10.1016/J.TRA.2005.06.007 Clark, B., Chatterjee, K., Melia, S., Knies, G., Laurie, H.: Life events and travel behavior exploring the interrelationship using UK Household Longitudinal Study data. Transp. Res. Rec. 2413 , 54–64 (2014). https://doi.org/10.3141/2413-06 Daly, A., Zachary, S.: Improved multiple choice models. In D. Hensher & O. Dalvi (Eds.), Identifying and Measuring the Determinants of Model Choice (pp. 187–201). (1979). https://www.researchgate.net/publication/230663926 de Haas, M.C., Kroesen, M., Chorus, C., Hoogendoorn-Lanser, S., Hoogendoorn, S.: E-bike user groups and substitution effects: evidence from longitudinal travel data in the Netherlands. Transportation. 49 (3), 815–840 (2022). https://doi.org/10.1007/S11116-021-10195-3/FIGURES/3 de Haas, M.C., Scheepers, C.E., Harms, L.W.J., Kroesen, M.: Travel pattern transitions: Applying latent transition analysis within the mobility biographies framework. Transp. Res. Part. A: Policy Pract. 107 , 140–151 (2018). https://doi.org/10.1016/j.tra.2017.11.007 Diana, M., Mokhtarian, P.L.: Grouping travelers on the basis of their different car and transit levels of use. Transportation. 36 (4), 455–467 (2009). https://doi.org/10.1007/s11116-009-9207-y Faber, R.M., Jonkeren, O., de Haas, M.C., Molin, E.J.E., Kroesen, M.: Inferring modality styles by revealing mode choice heterogeneity in response to weather conditions. Transp. Res. Part. A: Policy Pract. 162 , 282–295 (2022). https://doi.org/10.1016/J.TRA.2022.06.003 Fosgerau, M.: Investigating the distribution of the value of travel time savings. Transp. Res. Part. B: Methodological. 40 (8), 688–707 (2006). https://doi.org/10.1016/J.TRB.2005.09.007 Gao, J., He, S.Y., Ettema, D., Helbich, M.: Travel behavior changes due to life events: Longitudinal evidence from Dutch couple households. Transp. Res. Part. A: Policy Pract. 175 , 103765 (2023). https://doi.org/10.1016/J.TRA.2023.103765 Gärling, T., Axhausen, K.W.: Introduction: Habitual travel choice. Transportation. 30 (1), 1–11 (2003). https://doi.org/10.1023/A:1021230223001/METRICS Hamaker, E.L.: Why researchers should think within-person: A paradigmatic rationale RI-CLPM. In Handbook of Research Methods for Studying Daily life (pp. 43–61). (2012). https://www.researchgate.net/publication/266896375 Hensher, D.A., Greene, W.H.: The mixed logit model: The state of practice. Transportation. 30 (2), 133–176 (2003). https://doi.org/10.1023/A:1022558715350/METRICS Hess, S., Daly, A.J.: Handbook of choice modelling. Edw. Elgar publishing (2014). https://www.e-elgar.com/shop/gbp/handbook-of-choice-modelling-9781781003145.html Hess, S., Palma, D.: Apollo: a flexible, powerful and customisable freeware package for choice model estimation and application version 0.0.7 User manual. www.ApolloChoiceModelling.com (2019) Hess, S., Palma, D., Calastri, C., Crasted dit Sourd, R., Daly, A., Dumont, J., Molloy, J., Schmid, B.: Apollo: a flexible, powerful and customisable freeware package for choice model estimation and application. In Journal of Choice Modelling: Vol. In Press. Elsevier. (2019). https://doi.org/10.1016/J.JOCM.2019.100170 Hoogendoorn-Lanser, S., Schaap, N.T.W., Olde Kalter, M.J.: The netherlands mobility panel: An innovative design approach for web-based longitudinal travel data collection. Transp. Res. Procedia. 11 , 311–329 (2015). https://doi.org/10.1016/j.trpro.2015.12.027 Janke, J., Handy, S.: How life course events trigger changes in bicycling attitudes and behavior: Insights into causality. Travel Behav. Soc. 16 , 31–41 (2019). https://doi.org/10.1016/J.TBS.2019.03.004 Keskisaari, V., Ottelin, J., Heinonen, J.: Greenhouse gas impacts of different modality style classes using latent class travel behavior model. J. Transp. Geogr. 65 , 155–164 (2017). https://doi.org/10.1016/j.jtrangeo.2017.10.018 Kløckner, C.: How single events change travel mode choice: a life span perspective. (2013). https://ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/2392913 Kroesen, M.: Modeling the behavioral determinants of travel behavior: An application of latent transition analysis. Transp. Res. Part. A: Policy Pract. 65 , 56–67 (2014). https://doi.org/10.1016/j.tra.2014.04.010 Kroesen, M.: To what extent do e-bikes substitute travel by other modes? Evidence from the Netherlands. Transp. Res. Part. D: Transp. Environ. 53 , 377–387 (2017). https://doi.org/10.1016/j.trd.2017.04.036 Liao, F., Molin, E., Timmermans, H., van Wee, B.: The impact of business models on electric vehicle adoption: A latent transition analysis approach. Transp. Res. Part. A: Policy Pract. 116 , 531–546 (2018). https://doi.org/10.1016/J.TRA.2018.07.008 Loder, A., Axhausen, K.W.: Mobility tools and use: Accessibility’s role in Switzerland. J. Transp. Land. Use. 11 (1), 367–385 (2018). https://doi.org/10.5198/JTLU.2018.1054 Louviere, J., Hensher, D.A., Swait, J.D.: Stated choice methods: analysis and application. Cambridge University Press (2000). https://doi.org/10.1017/CBO9780511753831.008 McFadden, D.: Conditional logit analysis of qualitative choice behavior. Front. Econometrics, 105–142. (1974). http://elsa.berkeley.edu/reprints/mcfadden/zarembka.pdf McNally, M.G.: The Four Step Model, Handbook of Transport Modeling. In Institute of Transportation Studies and Department of Civil & Environmental Engineering. (2000). https://escholarship.org/uc/item/7j0003j0 Molin, E., Mokhtarian, P., Kroesen, M.: Multimodal travel groups and attitudes: A latent class cluster analysis of Dutch travelers. Transp. Res. Part. A: Policy Pract. 83 , 14–29 (2016). https://doi.org/10.1016/j.tra.2015.11.001 Müggenburg, H., Busch-Geertsema, A., Lanzendorf, M.: Mobility biographies: A review of achievements and challenges of the mobility biographies approach and a framework for further research. J. Transp. Geogr. 46 , 151–163 (2015). https://doi.org/10.1016/J.JTRANGEO.2015.06.004 Murphy, J.J., Allen, P.G., Stevens, T.H., Weatherhead, D.: A meta-analysis of hypothetical bias in stated preference valuation. Environ. Resource Econ. 30 (3), 313–325 (2005). https://doi.org/10.1007/S10640-004-3332-Z/METRICS Neal, D.T., Wood, W., Labrecque, J.S., Lally, P.: How do habits guide behavior? Perceived and actual triggers of habits in daily life. J. Exp. Soc. Psychol. 48 (2), 492–498 (2012). https://doi.org/10.1016/J.JESP.2011.10.011 Nurul Habib, K., Weiss, A., Hasnine, S.: On the heterogeneity and substitution patterns in mobility tool ownership choices of post-secondary students: The case of Toronto. (2018). https://doi.org/10.1016/j.tra.2018.06.002 Olde Kalter, M.J., Paix Puello, L., L., Geurs, K.T.: Exploring the relationship between life events, mode preferences and mode use of young adults: A 3-year cross-lagged panel analysis in the Netherlands. Travel Behav. Soc. 24 , 195–204 (2021). https://doi.org/10.1016/j.tbs.2021.04.004 Prato, C.G., Halldórsdóttir, K., Nielsen, O.A.: Latent lifestyle and mode choice decisions when travelling short distances. Transportation. 44 (6), 1343–1363 (2017). https://doi.org/10.1007/s11116-016-9703-9 Rau, H., Manton, R.: Life events and mobility milestones: Advances in mobility biography theory and research. (2016). https://doi.org/10.1016/j.jtrangeo.2016.02.010 Scheiner, J.: Mobility Biographies and Mobility Socialisation—New Approaches to an Old Research Field. Life-Oriented Behavioral Research for Urban Policy, 385–401. (2017). https://doi.org/10.1007/978-4-431-56472-0_13 Schwanen, T., Banister, D., Anable, J.: Rethinking habits and their role in behaviour change: the case of low-carbon mobility. J. Transp. Geogr. 24 , 522–532 (2012). https://doi.org/10.1016/J.JTRANGEO.2012.06.003 Scott, D.M., Axhausen, K.W.: Household mobility tool ownership: Modeling interactions between cars and season tickets. Transportation. 33 (4), 311–328 (2006). https://doi.org/10.1007/S11116-005-0328-7/METRICS Small, K.A.: Valuation of travel time. Econ. Transp. 1 (1–2), 2–14 (2012). https://doi.org/10.1016/J.ECOTRA.2012.09.002 Strömberg, H., Rexfelt, O., Karlsson, I.C.M.A., Sochor, J.: Trying on change – Trialability as a change moderator for sustainable travel behaviour. Travel Behav. Soc. 4 , 60–68 (2016). https://doi.org/10.1016/J.TBS.2016.01.002 Train, K.E.: Discrete choice methods with simulation, second edition. In Discrete Choice Methods with Simulation, Second Edition (Vol. 9780521766). (2009). https://doi.org/10.1017/CBO9780511805271 van Cranenburgh, S., Kouwenhoven, M.: An artificial neural network based method to uncover the value-of-travel-time distribution. Transportation. 48 (5), 2545–2583 (2021). https://doi.org/10.1007/S11116-020-10139-3/TABLES/9 Vij, A., Carrel, A., Walker, J.L.: Incorporating the influence of latent modal preferences on travel mode choice behavior. Transp. Res. Part. A: Policy Pract. 54 , 164–178 (2013). https://doi.org/10.1016/j.tra.2013.07.008 Wardman, M., Chintakayala, V.P.K., de Jong, G.: Values of travel time in Europe: Review and meta-analysis. Transp. Res. Part. A: Policy Pract. 94 , 93–111 (2016). https://doi.org/10.1016/J.TRA.2016.08.019 Wiggins, L.M.: Mathematical models for the interpretation of attitude and behavior change: The analysis of multi-wave panel. Columbia University ProQuest Dissertations & Theses. (1955) Williams, H.C.W.L.: On the Formation of Travel Demand Models and Economic Evaluation Measures of User Benefit. (1977). Http://Dx.Doi.Org/10.1068/A090285 , 9(3), 285–344. https://doi.org/10.1068/A090285 Xiong, C., Chen, X., He, X., Guo, W., Zhang, L.: The analysis of dynamic travel mode choice: a heterogeneous hidden Markov approach. Transportation. 42 (6), 985–1002 (2015). https://doi.org/10.1007/S11116-015-9658-2/TABLES/6 Zarwi, E., Vij, W., Zarwi, F., El, Vij, A., Walker, J.L.: Modeling and Forecasting the Evolution of Preferences over Time. A Hidden Markov Model of Travel Behavior (2017) Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5353959","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":375605810,"identity":"12e5f697-5759-4447-9166-24976ce54ac7","order_by":0,"name":"Roel Faber","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA40lEQVRIiWNgGAWjYHACxgMMNgwJDAzMB4CkBETsAQE9BxjSQFrYEhBaEojTwmOAEMKnxby9+QDQQXZ5/NJnvm54uMNCzlwige0BPi0yZ44lALUkF0v25W67kXhGwthyRgK7AT4tEhI5BgcYfzAnbjjDC9TSJpG44XYCmwR+LfkfgLbUJ+4/w/MMpKWeCC05QO8nHE7cwMPDBtKSYEBQC88xgwMJCceLJc6wmYG0GG64/7ANvxb25ocPPiRU5/H3MD+7+bOtTt7gzOFjEh/waAEDNDMZGwhpGAWjYBSMglFAAAAAARhQ4H2WsVEAAAAASUVORK5CYII=","orcid":"","institution":"Delft University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Roel","middleName":"","lastName":"Faber","suffix":""},{"id":375605811,"identity":"a73f7e92-b8cb-44f9-aec5-843ba73fbeaa","order_by":1,"name":"Sander van Cranenburgh","email":"","orcid":"","institution":"Delft University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Sander","middleName":"van","lastName":"Cranenburgh","suffix":""},{"id":375605812,"identity":"43d2ad8b-61de-4732-80ef-8ab56e3ca62a","order_by":2,"name":"Maarten Kroesen","email":"","orcid":"","institution":"Delft University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Maarten","middleName":"","lastName":"Kroesen","suffix":""},{"id":375605813,"identity":"5f0760f3-b584-46c2-856f-8bca460c10e6","order_by":3,"name":"Eric Molin","email":"","orcid":"","institution":"Delft University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Eric","middleName":"","lastName":"Molin","suffix":""}],"badges":[],"createdAt":"2024-10-29 11:38:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5353959/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5353959/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":68737629,"identity":"9a8da274-7cf3-4cd7-9358-3f3d492edcaa","added_by":"auto","created_at":"2024-11-11 13:50:09","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":31797,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual model of the latent transition choice model\u003c/p\u003e","description":"","filename":"floatimage164.png","url":"https://assets-eu.researchsquare.com/files/rs-5353959/v1/ac9343e21ed33626432b985f.png"},{"id":68737630,"identity":"96b221f2-e3f6-458b-b91b-56e138125628","added_by":"auto","created_at":"2024-11-11 13:50:09","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":103800,"visible":true,"origin":"","legend":"\u003cp\u003eEstimated conditional mode choice probabilities on average trips between 0 and 20 km\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5353959/v1/50c741407c7cb6f3933d0bea.png"},{"id":68738785,"identity":"c7960b8c-dd73-48fb-bec3-cf48f67ee896","added_by":"auto","created_at":"2024-11-11 13:58:09","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":119394,"visible":true,"origin":"","legend":"\u003cp\u003eEstimated mean mode choice probabilities conditional on (changes in) car ownership\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5353959/v1/79af23f1740508b13798088f.png"},{"id":68737632,"identity":"85d1c9a6-33ce-46da-9a7f-008e191ed4ee","added_by":"auto","created_at":"2024-11-11 13:50:09","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":94899,"visible":true,"origin":"","legend":"\u003cp\u003eEstimated choice probabilities, conditional on (changes in) e-bike ownership\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5353959/v1/0c632ab26e5224c448eed75d.png"},{"id":89107818,"identity":"522bb160-9194-4048-ad19-36b6c3cc9f61","added_by":"auto","created_at":"2025-08-14 18:01:34","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1860598,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5353959/v1/7c2a80f8-1821-491f-9467-5e5b67d8073c.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"The effects of life-events and changes in mobility tool ownership on mode choice behaviour","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eMode choice analysis is a fundamental subject within travel behaviour research (McNally, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2000\u003c/span\u003e). The dominant approach to modelling mode choice behaviour has been based on utility-maximization discrete choice theory since the 1970s (McFadden, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1974\u003c/span\u003e; Train, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Traditionally, discrete choice models are employed in a static fashion, meaning that they do not consider changes in preferences over time (Ben-Akiva et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). Given the habitual nature of travel behaviour (Aarts \u0026amp; Dijksterhuis, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; G\u0026auml;rling \u0026amp; Axhausen, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) this assumption seems reasonable. Habits can often be difficult to break, frustrating policymakers\u0026rsquo; efforts to change travel behaviour. But behavioural changes \u003cem\u003edo\u003c/em\u003e occasionally occur (Schwanen et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Str\u0026ouml;mberg et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Moreover, from a policy perspective, understanding when individuals change their travel behaviour and travel preferences is vital information for crafting policies to achieve desired behavioural changes. Such periods of change can be seen as \u0026lsquo;windows of opportunity\u0026rsquo;, where policies can have a larger impact on behavioural change.\u003c/p\u003e \u003cp\u003eOne research stream focusing on behavioural change is the mobility biographies framework, where the study of behavioural change has focused on life-events, such as residential relocations, starting families, or changing jobs (M\u0026uuml;ggenburg et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Rau \u0026amp; Manton, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). In this literature, these life-events are then seen as windows of opportunity for behavioural change (Scheiner, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Previous work has indeed shown that people\u0026rsquo;s habits are indeed prone to change during these events (de Haas et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; M\u0026uuml;ggenburg et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Aside from life-events, which are more general changes as one goes through life, research has also investigated the effects of mobility tool ownership, such as vehicle or public transport card ownership, on travel behaviour (Loder \u0026amp; Axhausen, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Nurul Habib et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Scott \u0026amp; Axhausen, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). However, these studies have only looked at the effect of life-events on mode use (Gao et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), rather than on mode choice. Mode use here is defined as the result of both the trip generation process (which trips will a certain person make) and the mode choice process (given a certain trip, which mode will a certain person choose). Since mode use combines both trip generation and mode choice, studies that focus on mode use, for example in terms of total kilometres or number of trips travelled with certain travel modes, cannot disentangle possible changes in trip-generation from changes in mode choice preferences. Furthermore, these studies typically use clustering approaches(de Haas et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Olde Kalter et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) as opposed to discrete choice modelling techniques. Consequently, they have not been able to study how attribute tastes and preferences change as a result of life-events.\u003c/p\u003e \u003cp\u003eIn a separate research stream emerging in the last decade, some transport researchers have developed longitudinal choice models to study mode choice behaviour (Xiong et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Zarwi et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). These longitudinal choice models enable the study of evolving preferences over time. However, this previous work however did not focus on the effects of multiple changes within the individual, such as life-events. Zarwi et al. (\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) studied the effects of changes to the transportation system, whereas Xiong et al. (\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) looked at one set of general \u0026lsquo;life-stages\u0026rsquo;.\u003c/p\u003e \u003cp\u003eThis paper intends to further combine the fields of mobility biographies, including mobility tool-ownership, and the longitudinal study of mode choice behaviour. The main aim is to study how individual characteristics, including life-events and changes in mobility tool ownership, shape mode choice preferences over time. To achieve this objective, we will employ the relatively rare latent transition choice model. This model is an extension of a normal latent class choice model, where time dynamics are explicitly considered in the class-membership function and individual respondents\u0026rsquo; transitions between classes over time are modelled explicitly. As a result, it is a relatively parsimonious way to effectively model the change of mode choice preferences over time. The model utilizes panel data from the Netherlands Mobility Panel (MPN), which is enriched with the life-events and changes in mobility tool ownership.\u003c/p\u003e"},{"header":"2. Literature overview and Conceptual Model","content":"\u003cp\u003eIn this section, we will introduce the relevant literature and use it to build a conceptual model that will guide the analyses in this paper. The literature overview consists of three building blocks: first, we will provide an overview of studies into behavioural change. Second, we will look at mode choice analysis and the idea of modality styles and third, we provide an overview of the literature on longitudinal choice models in travel behaviour research. Finally, we synthesize the findings from these building blocks in a conceptual model.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Behavioural change in transportation\u003c/h2\u003e \u003cp\u003eOne subject area within travel behaviour research is that of the inertia that is present when people make decisions. This area works under the assumption that decision makers\u0026rsquo; choices are driven to some extent by habits, contrasting the usual assumption of full rationality for each new choice situation that underlies typical choice models (Aarts \u0026amp; Dijksterhuis, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; G\u0026auml;rling \u0026amp; Axhausen, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Neal et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The formation and breakdown of these habits are key topics of interest. Moments when habits are broken down are seen as the key windows of opportunity for changing people\u0026rsquo;s travel behaviour. Two such potential \u0026lsquo;habit-breaking\u0026rsquo; moments are life-events and changes in mobility tool ownership (Clark et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Gao et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Janke \u0026amp; Handy, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLife-events are key events in one\u0026rsquo;s life course that entail a disruption to day-to-day life (de Haas et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; M\u0026uuml;ggenburg et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), therefore breaking habitual travel behaviour. Such events can be related to family life (childbirth, marriage, leaving the home, residential relocations) and employment (gaining or losing employment, gaining or losing working from home abilities; Gao et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). These life-events can prompt a person to re-evaluate their habitual behaviour and thus provide a critical window of opportunity to enact behavioural change (Janke \u0026amp; Handy, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Kl\u0026oslash;ckner, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Up to this point, the longitudinal studies on life-events in travel behaviour research has focused on cluster analyses (de Haas et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Kroesen, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). These analyses reveal shifts in travel patterns, but they are unable to show changes in attribute preferences and elasticities as a result of the life-event.\u003c/p\u003e \u003cp\u003eAnother area of travel behaviour research focuses on so-called mobility tools: vehicles, drivers\u0026rsquo; licenses, and (discount) passes for public transport (Scott \u0026amp; Axhausen, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). The acquisition of these tools is based on longer-run expectations of mobility needs. Simultaneously, these tools allow their owners to more easily, cheaply, and/or effectively use certain transport modes. As a result, they create lock-in effects (Scott \u0026amp; Axhausen, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2006\u003c/span\u003e): owning a car makes the car a more attractive option even for trips that might be more suitable to the bicycle or public transport. Consequently, mobility tool ownership is closely linked to the existence of travel habits and could be a crucial explanation for the modality styles introduced above.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Mode Choice analysis and modality styles\u003c/h2\u003e \u003cp\u003eDiscrete choice modeling has been a cornerstone of travel behaviour research since the introduction of random utility-maximization theory within discrete choice analysis (McFadden, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1974\u003c/span\u003e). Two impactful applications of these models are the estimation of the value of travel time, typically using stated preference data based on route choices (Small, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Wardman et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), and modeling of mode choice in large-scale travel demand models (Ben-Akiva \u0026amp; Lerman, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e1985\u003c/span\u003e; McNally, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Train, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). One key improvement on the discrete choice theory workhorse, the multinomial logit model, is that of the nested logit model (Carrasco et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Daly \u0026amp; Zachary, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1979\u003c/span\u003e; Williams, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e1977\u003c/span\u003e). The nested logit model groups subsets of alternatives which are similar in some unobserved characteristics, which enables it to remove the irrelevance of independent alternatives (IIA) property of the multinomial logit model. This improvement is highly relevant in the case of mode choice analysis, as some modes compete more with other similar modes than they do others (Train, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOwing to the exponential increase in computing resources, the investigation of preference-heterogeneity within the population rose to the forefront of modeling efforts since the 1990s. For example, much interest has been paid to the distribution of the value of travel time across the population (Cirillo \u0026amp; Axhausen, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Fosgerau, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Hensher \u0026amp; Greene, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; van Cranenburgh \u0026amp; Kouwenhoven, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Simultaneously, for mode choice analysis, research has shown how preferences for various modes vary across the population (Bhat, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Cherchi et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). One concept that helps to communicate heterogeneity in mode choice preferences is that of modality styles. Modality styles are discrete segmentations, based on variations between people in underlying preferences to use certain modes (Diana \u0026amp; Mokhtarian, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Molin et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Vij et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Typically, these modality styles are uncovered using clustering methods, where people are grouped based on the number of times they make use of certain travel modes (Faber et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Another method to identify these modality styles is the latent class choice model (Faber et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Keskisaari et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Prato et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Vij et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Using this method has the advantage that it ties the concept of modality styles into discrete choice theory.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Longitudinal Choice Modeling and Latent Transition Choice Model\u003c/h2\u003e \u003cp\u003eTraditionally, choice models have been employed in a static, time-indifferent, fashion (Train, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). These models are agnostic to the process or order of the choices made and therefore operates under the implicit assumption that (from the modellers\u0026rsquo; perspective) all choices are made more or less simultaneously. As a result, the time when a choice is made is not considered in the modelling process. We want to note that for many, perhaps even the vast majority of choice modelling, this assumption is completely valid (Ben-Akiva et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). However, this static paradigm does entail that the analyst is unable to determine the effects of changes in characteristics, either on the level of the trip or the decision-maker, on the choice probabilities over time (Hamaker, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Instead, differences \u003cem\u003ebetween\u003c/em\u003e respondents can be used to estimate the effects of certain characteristics. These effects are then often assumed to be similar to longitudinal changes \u003cem\u003ewithin\u003c/em\u003e respondents over time, for example when the choice model is used to forecast future travel demand after some changes have been made. The problem with this method is that it is uncertain whether the \u003cem\u003ebetween-\u003c/em\u003erespondent effects are similar to the effects \u003cem\u003ewithin\u003c/em\u003e respondents. Take for example the potential effect of electric bicycle ownership. Within a static approach, a model might estimate the effect of e-bike ownership on bicycle choice probability by comparing the difference in choice probabilities across respondents that either do or do not own such a vehicle. However, it is not unlikely that respondents who own an electric bicycle are more avid cyclists in the first place, which prompted them to buy an e-bike. There is thus likely to be a rather strong self-selection effect, making it difficult to ascertain the effect that electric bicycle ownership would have on the choice behaviour of people who do not yet own such a vehicle.\u003c/p\u003e \u003cp\u003eTwo potential approaches to solve this problem are commonly found in the literature: first, and most prevalent, is the use of stated preference data. These analyses explicitly ask respondents to make choices considering hypothetical scenarios, where the researcher is free to design the experiment, and they can vary the attributes or scenarios associated with certain choices. In our example, respondents could be prompted to make choices between the bicycle and the car, first without considering electric bicycle ownership and then in the hypothetical scenario where they did own such a vehicle. However, as stated preference choices are made \u0026lsquo;on paper\u0026rsquo; they suffer from a potential lack of external validity (Louviere et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Murphy et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). For example, respondents might not be familiar enough with an electronic bicycle to give reliable estimates of how owning such a vehicle would change their mode choice behaviour.\u003c/p\u003e \u003cp\u003eThe second solution is to use multiple measurements per individual, ideally taken across a wide enough range of time, and to then explicitly model the time when the choices are made within the modelling context (Ben-Akiva et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). One example of such a choice model is the latent transition choice model, also known as the Markov choice model (Ben-Akiva et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Liao et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Xiong et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Zarwi et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). The latter name derives from the Markov process, where the probability of each event only depends on the state of the previous event and not on any previous states. The former name, which we prefer to use in this article as it aligns more closely with common nomenclature in the field of travel behaviour research, originates from latent transition cluster modeling, which has been a popular approach to study longitudinal data in the field of travel behaviour research for some time (see for example Kroesen, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The model works by allowing the class membership probabilities, estimated on the level of the individual, to change over time (Anderson, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1954\u003c/span\u003e; B\u0026ouml;ckenholt \u0026amp; Langeheine, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e1996\u003c/span\u003e; Wiggins, \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e1955\u003c/span\u003e). These changes, or transitions, can then be affected by external factors or other changes within the individual, such as life-events or changes in mobility tool ownership. As a result, these models allow for the modeller to estimate the effects of changing circumstances or explanators of choice behaviour on the level of the individual. It therefore allows us to estimate the effect of certain characteristics on choice behaviour \u003cem\u003ewithin\u003c/em\u003e individuals.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Conceptual Model\u003c/h2\u003e \u003cp\u003eThe relationships in the literature mentioned above are graphically summarized below in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. This conceptual model is then used to further guide the analyses in the paper.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo build the conceptual model, we start with the discrete choice building block, which assumes that observed mode choices can be explained using both the observed and unobserved attributes of each travel mode alternative for the trip. In this study, we will use alternative specific travel times and travel distances as these attributes. We then assume that distinct modality styles exist, and we let the effect of the attributes on mode choice vary across these modality styles. Then we allow socio-demographic factors, life-events, and mobility tool ownership to affect the class membership probability of these modality styles. Following the reasoning that the modality styles are largely inert, class membership in the first wave affects class membership in the second wave. Finally, we include a moderating effect of life-events, (changes in) mobility tool ownership, and socio-demographic factors on the probability that someone switches between modality styles.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Research methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Mathematical Model\u003c/h2\u003e \u003cp\u003eThe latent transition choice model builds on the standard latent class choice model, with the addition of latent transition parameters, which are used to estimate the class transition probabilities between the longitudinal waves. For an overview of the mathematical definition of the latent class choice model, the reader is referred to Ben-Akiva et al. (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1997\u003c/span\u003e) and Hess \u0026amp; Daly (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor the latent transition choice model, the log-likelihood function for observing a series of choices \u003cem\u003ek\u003c/em\u003e made by decision maker \u003cem\u003en\u003c/em\u003e belonging to class \u003cem\u003es\u003c/em\u003e at timepoint \u003cem\u003et\u003c/em\u003e, with alternatives \u003cem\u003ei\u003c/em\u003e, can be written as a function of taste parameters \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{s}\\)\u003c/span\u003e\u003c/span\u003e. The taste parameters are thus conditional on decision maker \u003cem\u003en\u003c/em\u003e belonging to class \u003cem\u003es\u003c/em\u003e at wave \u003cem\u003et\u003c/em\u003e in Eq.\u0026nbsp;1:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:LL\\left(\\beta\\:\\right)=\\sum\\:_{n\\:=\\:1}^{N}ln\\prod\\:_{t\\:=\\:1}^{T}\\sum\\:_{s\\:=1}^{S}{\\pi\\:}_{n,t,s}\\left(\\prod\\:_{k\\:=\\:1}^{K}{P}_{n,t}\\left({i}_{k,t}|{\\beta\\:}_{s}\\right)\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e( 1 )\u003c/p\u003e \u003cp\u003eThe key addition here is that the class-membership probability \u003cem\u003eπ\u003c/em\u003e is dependent on the wave \u003cem\u003et\u003c/em\u003e at which the choice was made. The class-membership function of the latent classes for wave 1 follows conventional standards and is estimated as a multinomial logit function based on an initial class-specific constant \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\delta\\:}_{s}\\)\u003c/span\u003e\u003c/span\u003e, as well as a function \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:g\\)\u003c/span\u003e\u003c/span\u003e of a vector of parameters \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{y}_{s}\\)\u003c/span\u003e\u003c/span\u003e and a vector of socio-demographic characteristics, life-events, and (changes in) mobility tool ownership \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{z}_{n,t}\\)\u003c/span\u003e\u003c/span\u003e, as given in Eq.\u0026nbsp;2.\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:{\\pi\\:}_{n,s,t\\:=\\:1}\\:=\\:\\frac{exp({\\delta\\:}_{s}\\:+\\:g({y}_{s},\\:{z}_{n,t}\\left)\\right)}{\\sum\\:_{s}exp({\\delta\\:}_{s}\\:+\\:g({y}_{s},\\:{z}_{n,t}\\left)\\right)}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e( 2 )\u003c/p\u003e \u003cp\u003eThe class-membership function of the latent classes for the second wave, however, is specified to be conditional on the class-membership probability of the first wave. The transitions between the classes are then modelled as in Eq.\u0026nbsp;3, where the probability that a decision maker \u003cem\u003en\u003c/em\u003e who belonged to class \u003cem\u003er\u003c/em\u003e in wave 1 will belong to class \u003cem\u003es\u003c/em\u003e at wave 2 is equal to the transition probability \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{tr}_{n,s,r}\\)\u003c/span\u003e\u003c/span\u003e of class \u003cem\u003er\u003c/em\u003e to class \u003cem\u003es\u003c/em\u003e, multiplied by the class-membership probability of belonging to class \u003cem\u003er\u003c/em\u003e in wave 1:\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:{\\pi\\:}_{n,s,t\\:=\\:2}\\:=\\sum\\:_{s=1}^{S}{\\sum\\:}_{r=1}^{R}{\\:(tr}_{n,s,r})\\:\\left({\\pi\\:}_{n,t\\:=\\:1,\\:\\:r}\\right)\\:\\:\\:$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e( 3 )\u003c/p\u003e \u003cp\u003eThese transition probabilities themselves are modelled as multinomial logit functions as well, such that the transition probability depends on a transition parameter \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{{\\phi\\:}}_{s,r}\\)\u003c/span\u003e\u003c/span\u003e associated with the transition from class \u003cem\u003er\u003c/em\u003e at the previous wave to class \u003cem\u003es\u003c/em\u003e at the current wave and a function \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:g\\)\u003c/span\u003e\u003c/span\u003e of both parameters \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{y}_{s,r}\\)\u003c/span\u003e\u003c/span\u003e and a vector of sociodemographic characteristics, life-events and (changes in) mobility tool ownership \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{z}_{n,t}\\)\u003c/span\u003e\u003c/span\u003e, as given below in Eq.\u0026nbsp;4:\u003cdiv id=\"Equd\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equd\" name=\"EquationSource\"\u003e\n$$\\:{tr}_{n,s,r}=\\:\\frac{exp({{\\phi\\:}}_{s,r}\\:+\\:g({y}_{s,r},\\:{z}_{n,t}\\left)\\right)}{\\sum\\:_{s}exp({{\\phi\\:}}_{s,r}\\:+\\:g({y}_{s,r},\\:{z}_{n,t}\\left)\\right)}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e( 4 )\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Research Data\u003c/h2\u003e \u003cp\u003eWe use trip data from the travel diary of the Netherlands Mobility Panel ([MPN], for more information see Hoogendoorn-Lanser et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), a household panel in the Netherlands that comprises an extensive questionnaire and a 3-day travel survey. Respondents for the MPN are recruited from the Kantar NIPObase, an invite-only internet access panel (IAP). Invitations for the Kantar NIPObase are sent out based on register data. Members of the larger NIPObase IAP are then invited for the MPN separately, based on their socio-demographic characteristics. Between 30 and 50% of respondents from the larger IAP decide to join the MPN upon receiving an invitation. When respondents have entered the MPN, their yearly response rates for each wave vary around 85%.\u003c/p\u003e \u003cp\u003eWe use data from the yearly waves between 2016 and 2022. For each unique respondent, we select one set of two consecutive waves. If there were sets of consecutive waves where life-events or changes in mobility tool ownership happened between the two waves, then we always selected one of these sets. This procedure ensures that the final dataset contains as many life-events and changes in mobility tool ownership as possible. If no life-events or changes in mobility tool ownership happened, then one set of consecutive waves is drawn at random for each person. The sample descriptives for the final sample, as collected during the first wave used in the dataset, is given in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSample descriptives compared to population distribution\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSample\u003c/p\u003e \u003cp\u003e(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePopulation\u003c/p\u003e \u003cp\u003e(12\u0026thinsp;+\u0026thinsp;inhabitants of the Netherlands, 2019; %)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eAge (Years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12\u0026ndash;24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25\u0026ndash;44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e28\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e45\u0026ndash;64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e33\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65 +\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eEducation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLow\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMedium\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e26\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003eUrban Density residential municipality\u003c/p\u003e \u003cp\u003e(addresses/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e500\u0026ndash;1000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1000\u0026ndash;1500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1500\u0026ndash;2500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026gt;\u0026thinsp;2500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eHousehold Type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSingle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOnly adults\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e46\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdults and children\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe sample descriptives are very similar to the population values for nearly all variables. The only exception is household type, as our sample consists of comparatively more households with children. This is most likely the result of the biased sampling procedure introduced above, where we purposefully oversampled sets of waves that include life-events. As life-events are more common within households with children, these types of households will be oversampled as well. This enables us to identify the effects of life-events on the transition probabilities more reliably.\u003c/p\u003e \u003cp\u003eThe primary unit of analysis of the mode choice model is the trip. Principally, we analyse trips as recorded by respondents in each waves\u0026rsquo; travel diary. However, not all trips in the travel diary are useful for our analysis. Therefore, some selection criteria are used. First, only trips that departed from the residential location were selected, as the residence is typically the location where the mode choice decision is made. Second, all trips made with modes other than the car, public transport, the bicycle, or on foot were discarded. The shares of the other modes are marginal, and estimating valid attribute-parameters for them is therefore not feasible. Third, trips for which a very large distance (\u0026gt;\u0026thinsp;200km) was reported are excluded, as the decision-making process for such trips differs from that of the more typical, daily trips. These selection criteria leave us with a total number of 28,117 trips made by 4,789 unique respondents.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Operationalisation\u003c/h2\u003e \u003cp\u003eWe use alternative-specific travel times and travel distances as the trip-specific explanatory variables in the model. These travel times and -distances are calculated using the Google Directions API based on the origin, destination, and departure time of the trip. In the utility function, we use both a linear and a square root transformation of travel time for each mode to capture possible non-linear effects of travel time on the utility of each mode. Travel distance is used in the utility function of the active modes. Finally, aside from alternative specific constants for all trips we also use a dummy-variable for trips that are made with other people. This dummy variable is used to correct for the fact that such shared trips are more often made with the car compared to the other modes, allowing for better estimates of the travel time and travel distance parameters. This dummy variable is kept fixed across the latent classes. The final utility functions then are given in Eq.\u0026nbsp;5 below. Note that the parameter pertaining to travel distance is only estimated for the active modes.\u003cdiv id=\"Eque\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Eque\" name=\"EquationSource\"\u003e\n$$\\:Utility=\\:asc+dumm{y}_{shared\\:trip}+\\:{\\beta\\:}_{travel\\:time}+\\:{\\beta\\:}_{\\sqrt{travel\\:time}}+{\\beta\\:}_{travel\\:distance}\\:$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e( 5 )\u003c/p\u003e \u003cp\u003eThe utility functions are kept relatively simple, to balance with the complexity of estimating the transitions between latent classes. Estimating both a very complex utility function and the transition parameters would quickly lead to an over-specified model. For the travel times and travel distances, which are the key explanatory variables in the utility functions, descriptive statistics are given in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive statistics for the alternative-specific travel times and travel distances of all trips in the dataset\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMin.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMode\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMean.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMax.\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eTravel Time (min.)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e141\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePublic Transport\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e34.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2234\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e31.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e671\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWalking\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e33\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e114\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2373\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eTravel\u003c/p\u003e \u003cp\u003eDistance (km)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e212\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWalking\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e195\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn the class-membership and transition functions, we use socio-demographic characteristics, life-events, and (changes in) mobility tool ownership. For the life-events and (changes in) mobility tool ownership, we have used the variables given in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, which are presented together with their absolute and relative occurrence in the sample. The correlations between these variables are not very high, with the exception of the variables pertaining to car ownership and access to cars. The highest correlation here however is 0.68, and therefore there are no strong concerns regarding multicollinearity.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eList of life-events and mobility tool ownership variables and their occurrence in the data\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOccurrence (N; %)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eLife-events\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResidential relocation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e290 (6.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChange of job\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e558 (12%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBirth of a child\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e147 (3.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eShift to working from home\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e204 (4.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMobility Tool Ownership\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCar ownership\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,405 (71%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlways has access to car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,910 (61%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNever has access to car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e915 (19%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eE-bike ownership\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e955 (20%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOwns Personal Public Transport Card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,282 (69%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eChanges in Mobility Tool Ownership\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained personal car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e260 (5.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost personal car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e223 (4.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained access to car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e303 (6.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost access to car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e220 (4.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained electric bicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e515 (11%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained public transport card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e181 (3.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost public transport card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e175 (3.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cp\u003eThis section presents and discusses the main results. First, we show the goodness-of-fit of increasingly complex models, starting with a multinomial logit model and ending with a latent transition choice model which includes covariates. We show that the latter model provides the best fit to the data. Then, the class-specific parameters of this model are introduced, and we show that the latent classes can be interpreted as modality styles, as they reflect underlying predispositions to use certain travel modes. We will then discuss the class-membership results, which is followed by a discussion of the transitions between the two classes and the substantive meaning of these transitions.\u003c/p\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Model Selection\u003c/h2\u003e \u003cp\u003eThe choice models are primarily estimated in Apollo (Hess et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Hess \u0026amp; Palma, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). As the model is non-trivial to implement in Apollo, we validated the results by using a separate implementation in Matlab. To test whether the latent transition choice model offers an empirical benefit over more parsimonious models, we estimated six models, each increasing in complexity. Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e contains the goodness-of-fit statistics of these six models:\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003e1) A MNL model\u003c/h3\u003e\n\n\u003ch3\u003e2) A nested logit model\u003c/h3\u003e\n\n\u003ch3\u003e3) A conventional latent class choice model\u003c/h3\u003e\n \u003ch3\u003e4) A latent class choice model that allows for the sizes of the classes to vary per wave\u003c/h3\u003e\n\u003ch3\u003e5) A latent transition choice model\u003c/h3\u003e\n\n\u003ch3\u003e6) A latent transition choice model with covariates\u003c/h3\u003e\n\u003cp\u003eFrom the second model onwards, a nesting structure is used, with one nesting level. The root level contains the car and one nest, which contains the alternatives public transport, the bicycle, and walking. This nesting structure is based on the idea that public transport, the bicycle, and walking substitute each other more directly than they do the car. A comparison of the second model with the first model reveals whether this nesting structure improves model fit. Based on the literature and the conceptual model, we expect this to be the case for mode choice analysis on our revealed preference dataset. A more complicated nesting structure was tested where the active modes were separated into a further subnest, but this structure offered no improvement.\u003c/p\u003e \u003cp\u003eThe third model explores whether people\u0026rsquo;s preferences for mode alternatives are heterogeneous by estimating two separate latent classes. For the sake of parsimony in what is otherwise already a relatively complex model, we decided to fix the number of classes to two for all latent class models. Adding additional latent classes would result in an exponential growth of possible transitions and, therefore, transition parameters. Given the relatively modest probability of transitions between classes, we think that the dataset is not large enough to support more than two latent classes.\u003c/p\u003e \u003cp\u003eThe fourth model can capture behavioural changes across the population, which would result in different class sizes for the two waves. However, this model is not able to estimate which individuals\u0026rsquo; behaviour has changed. The fifth model, the latent transition choice model, is an improvement in that respect, as it can now assess which individuals transition between the estimated latent classes. Then, finally, we report a model (model 6) with covariates of both the initial class-membership and transition probabilities. This model enables an estimation of factors that influence the transition probabilities. These allow for much richer behavioural interpretations of the results.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGoodness-of-fit statistics for the choice models\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eModel 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModel 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eModel 3\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eModel 4\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eModel 5\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eModel 6\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMNL-model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNested Logit model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 class latent class model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2 class latent class model. change size across waves\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2 class latent transition choice model.\u003c/p\u003e \u003cp\u003eNo covariates\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2 class latent transition choice model.\u003c/p\u003e \u003cp\u003eWith covariates\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEst. parameters\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e \u003cp\u003e\u003cb\u003eGoodness-of-fit statistics\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN (individuals)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e \u003cp\u003e4 789\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN (choices)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e \u003cp\u003e28 117\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLL\u003csub\u003e0\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e \u003cp\u003e-37 923\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLL\u003csub\u003eβ\u003c/sub\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-20 421\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-19 941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-17 350\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-17 350\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-17 279\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-16 682\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean LL\u003csub\u003eβ\u003c/sub\u003e per person\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.724\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.706\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.617\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.617\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.616\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0.586\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eρ\u003csup\u003e2\u003c/sup\u003e eq.\u0026nbsp;shares\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.462\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.474\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.543\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.543\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.544\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.560\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eρ\u003csup\u003e2\u003c/sup\u003e obs. shares\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.314\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.331\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.418\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.418\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.420\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.440\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAIC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e40 863\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e39 904\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34 740\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e34 742\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e33 937\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e33 525\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBIC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e40 946\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e39 994\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34 905\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e34 915\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e34 784\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e34 185\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e \u003cp\u003e\u003cb\u003eCross-validation\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLL\u003csub\u003eβ\u003c/sub\u003e per obs.\u003c/p\u003e \u003cp\u003eIn sample\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.731\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.715\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.622\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.622\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.619\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0.598\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLL\u003csub\u003eβ\u003c/sub\u003e per obs.\u003c/p\u003e \u003cp\u003eOut of sample\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.706\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.689\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.597\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0.578\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e% Diff.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-3.57%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-3.61%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-3.52%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-3.52%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-3.53%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-3.37%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAs can be seen in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, the nested logit model (model 2) provides a much better fit to the data than the MNL model 1, indicating that there are nesting structures in the mode choice data. Furthermore, the nesting parameter lambda, is found to be significantly different from 1 (not reported in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), which also indicates the presence of nesting structures. The latent class models, starting with model 3, outperform the standard models, indicating the presence of mode choice heterogeneity. Model 4 does not provide a better fit to the data than model 3, meaning there are no substantial changes in mode choice behaviour across the population between the two waves. The transition model, model 5, however, provides a statistically significant better fit than both previous latent class models (LRT\u0026thinsp;=\u0026thinsp;142, df\u0026thinsp;=\u0026thinsp;1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). It also performs better in the 5-fold cross-validation tests, where the dataset is split into five parts and the model is subsequently estimated using four parts and tested on the remaining hold-out part. The model that uses covariates to determine the class-membership and transition probabilities (model 6) then statistically outperforms the latent class transition model without covariates (model 5) as well (LRT\u0026thinsp;=\u0026thinsp;1193, df\u0026thinsp;=\u0026thinsp;58, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This model will be used for further examination in the remainder of this paper.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Interpreting the latent classes as modality styles\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e reports the class-specific parameter estimates. Note that these parameter estimates, which are conditional on class-membership, are stable over time and thus the same for both waves.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEstimated class-specific model parameters\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eClass 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003eClass 2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCar\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePT\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBicycle\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eWalking\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eCar\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePT\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eBicycle\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eWalking\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eASC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-1.77\u003c/p\u003e \u003cp\u003e(-9.15)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003cp\u003e(5.16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.83\u003c/p\u003e \u003cp\u003e(18.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.22\u003c/p\u003e \u003cp\u003e(9.65)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3.67\u003c/p\u003e \u003cp\u003e(22.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e5.23\u003c/p\u003e \u003cp\u003e(13.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eShared Trip dummy\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.98\u003c/p\u003e \u003cp\u003e(-10.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.89\u003c/p\u003e \u003cp\u003e(-13.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.322\u003c/p\u003e \u003cp\u003e(-4.39)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003eSame as class 1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTravel time\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003e0.0176 (9.04)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003e0.026 (7.27)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSquare root travel time\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003e-0.910 (-10.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003e-0.86 (-10.05)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSquare root travel distance\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.389\u003c/p\u003e \u003cp\u003e(-10.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.996\u003c/p\u003e \u003cp\u003e(-11.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e-0.220\u003c/p\u003e \u003cp\u003e(-13.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e-0.712\u003c/p\u003e \u003cp\u003e(-6.50)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eNesting parameter\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003e0.725 (-7.45\u003csup\u003ea\u003c/sup\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003e0.493 (-12.0\u003csup\u003ea\u003c/sup\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"9\" nameend=\"c9\" namest=\"c1\"\u003e \u003cp\u003eIn this and following tables, robust t-ratios given between parentheses\u003c/p\u003e \u003cp\u003ea: Robust T-ratio with respect to 1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe find that the alternative-specific constants (ASCs) are very different for the two classes, which supports the notion that both classes represent modality styles that reflect underlying predispositions towards the use of certain travel modes. The ASCs of public transport, bicycle, and walking are all smaller for the first than the second class. This provides a first indication that the first class is in general more inclined towards the use of the car than the second class. Both travel time parameters are positive, but both square root travel time parameters are negative. The net effect of these parameters is that all modes\u0026rsquo; elasticities with respect to travel time are negative for both classes, as is to be expected. Similarly, travel distance has a negative effect on active mode use.\u003c/p\u003e \u003cp\u003eTo provide a more intuitive picture of the mode choice differences between the two classes, we have calculated the conditional probabilities for each of the two classes for trips falling within several \u0026lsquo;distance\u0026rsquo; bands. To do so, we used the average travel time per mode for all trips within the dataset that corresponded to a given distance. Note that this procedure introduces some noise, particularly for the longer distances, which average out travel times and travel distances over fewer trips. As a result, there are some fluctuations in the graph for distances lager than 10km. The estimated conditional probabilities for both classes and distances between 0 and 20 km are given in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows that there is a large difference in conditional choice probabilities between the two classes. For the first class, the probability of using the car increases rapidly as distances increase. The car becomes the dominant mode around 2.5km and then the choice probability asymptotically increases towards a choice probability nearing 1 for trips longer than roughly 10 km. For the second class however, the bicycle is the dominant mode up until trips of roughly 10 km in length. At that point the car becomes the most probable mode, followed by public transport. Based on these conditional mode choice probabilities for each class, we can identify the first class as \u0026lsquo;Car-oriented\u0026rsquo; and the second class as \u0026lsquo;Multimodal\u0026rsquo;. In the estimated choice model, roughly 60% of people belong to the \u0026lsquo;car-oriented\u0026rsquo; class and 40% to the \u0026lsquo;multimodal\u0026rsquo; class in the first wave.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.3. Class-membership function\u003c/h2\u003e \u003cp\u003eNow that the latent classes have been identified as two distinct modality styles (\u0026lsquo;car-oriented\u0026rsquo; and \u0026lsquo;multimodal\u0026rsquo;), we interpret the class-membership function. The parameters estimated in the class-membership function show the relation between class-membership and socio-demographics, life-events, and (changes in) mobility tool ownership. These parameters are given in Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. Importantly, the class-membership function is specific to membership in the first wave, before any transitions might have happened. The life-events and changes in mobility tool ownership have thus not happened yet. The direction of the causal effect therefore is not clear: either people with certain behaviours are more likely to undergo life-events/changes in mobility tool ownership or people might already be aware of upcoming life-events and have already changed behaviour accordingly. Thus, there might be both selection effects (first option) or lead-effects (second option).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eParameter estimates of the class-membership function for the first wave\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClass 2:\u003c/p\u003e \u003cp\u003e\u0026lsquo;Multi-modal\u0026rsquo;\u003c/p\u003e \u003cp\u003e(ref: class 1, \u0026lsquo;Car-\u003c/p\u003e \u003cp\u003eoriented\u0026rsquo;)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003edelta\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.06 (-2.36)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSocio-demographics\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;24\u003c/p\u003e \u003cp\u003e(ref: age 65+)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.11 (5.13)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge 25\u0026ndash;44\u003c/p\u003e \u003cp\u003e(ref. age 65+)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.0293 (-0.163)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge 45\u0026ndash;64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.129 (-0.841)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEmployed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0828 (0.0653)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWorks from home\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.05 (-0.439)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLow income\u003c/p\u003e \u003cp\u003e(ref: med income)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.269 (2.64)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHigh income (ref: med income)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.172 (1.15)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChildren in household\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0790 (0.65)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUrban Density\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.0547 (-0.84)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLife-events\u003c/b\u003e\u003c/p\u003e \u003cp\u003e(between wave 1 and wave 2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNew Job\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.124 (-0.86)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStarted working from home\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0793 (0.363)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChild Born\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.564 (-2.20)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResidential Relocation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.063 (0.33)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMobility Tool Ownership\u003c/b\u003e, wave 1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePersonal Car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.12 (-6.02)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccess to Car, always\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.91 (-13.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccess to Car, never\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.443 (1.40)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDrivers\u0026rsquo; License\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-2.01 (-5.50)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eElectronic Bicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.938 (7.84)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePT-card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.396 (3.69)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eChanges in Mobility Tool Ownership\u003c/b\u003e\u003c/p\u003e \u003cp\u003e(between wave 1 and wave 2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost personal car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.363 (1.84)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained personal car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.676 (-2.80)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost access to car\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.33 (6.95)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained access to car\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.892 (-4.06)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained electric bicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.840 (5.88)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained PT-card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.395 (1.85)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost PT-card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.42 (-1.51)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe socio-demographic and mobility tool ownership parameters show that the class-membership estimates are congruent with the earlier identification of the classes as modality styles. People who own a car, a driver\u0026rsquo;s license, and always have access to a car are more likely to belong to the car-oriented modality style. Conversely, people who own electronic bicycles or public transportation cards are more likely to belong to the multi-modal modality style. Similarly, as expected, young people and people with lower incomes are more likely to belong to the multi-modal modality style as well.\u003c/p\u003e \u003cp\u003eFurthermore, the class-membership estimates reveal some interesting effects of life-events and changes in mobility tool ownership on modality style membership. Prospective parents were more likely to belong to the \u0026lsquo;car-oriented\u0026rsquo; modality style before the child was born, even after controlling for the effects of age. Similar lead- or selection effects can be found for most changes in mobility tool ownership. People who bought a personal car between waves 1 and 2 were already more likely to belong to the \u0026lsquo;car-oriented\u0026rsquo; modality style than would otherwise be expected, as were people who gained the ability to always access a car. A similar but opposite effect is found for the addition of an electric bicycle: people who gained an electric bicycle in between the two waves were already more likely to belong to the multi-modal modality style.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Transitions between modality styles\u003c/h2\u003e \u003cp\u003eThe main advantage of the latent transition choice model is that we can combine the above within-class mode choice probabilities with calculations regarding the transition probabilities between the latent classes. Furthermore, the model calculates to which extent the transition probabilities between the classes are affected by life-events and changes in mobility tool ownerships.\u003c/p\u003e \u003cp\u003eBelow, we first present the statistically significant transition parameters and the socio-demographic, life-event, and mobility tool ownership parameters in Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e. For the sake of parsimony, variables without any significant effect are not shown in the table. Following a short discussion of the existence and direction of some effects below the table, we will illustrate the effects of a selection of life events and changes in mobility tool ownership on transition probabilities.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eParameter estimates of the transitions (non-significant parameters not shown)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTransition\u003c/p\u003e \u003cp\u003e\u0026lsquo;Car-oriented\u0026rsquo; to\u003c/p\u003e \u003cp\u003e\u0026lsquo;Multi-modal\u0026rsquo;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTransition\u003c/p\u003e \u003cp\u003e\u0026lsquo;Multi-modal\u0026rsquo; to\u003c/p\u003e \u003cp\u003e\u0026lsquo;Car-oriented\u0026rsquo;\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTransition parameter (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}\\)\u003c/span\u003e\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-7.84 (-3.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-5.50 (-4.53)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSocio-demographics\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u0026thinsp;\u0026lt;\u0026thinsp;=\u0026thinsp;24\u003c/p\u003e \u003cp\u003e(ref: age 25+)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.50 (2.80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLife-events\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNew Job\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.785\u003csup\u003ea\u003c/sup\u003e (1.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStarted Working from Home\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.66 (2.05)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHigher Income\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.91 (2.06)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResidential Relocation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.84\u003csup\u003ea\u003c/sup\u003e (1.77)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMobility Tool Ownership in first wave\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccess to Car, always\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.08 (-2.38)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccess to Car, never\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-1.32 (-1.97)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eElectronic Bicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.26 (3.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePT-card\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.71 (2.84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eChanges in Mobility Tool Ownership\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained personal car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.51 (2.05)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained access to car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.84 (2.63)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGained electric bicycle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.76 (4.35)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003ea: only statistically significant if we accept a 10% threshold for significance testing\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eBoth transition parameters are negative and statistically significant, indicating that transitions between classes are relatively rare events. There is a very limited effect of static socio-demographic variables on the transition probabilities. This is to be expected, given that the effects of static socio-demographic covariates on class membership are already estimated in the class-membership model of the first wave. There are some effects of life-events, although they are not very strong. Interestingly, both working from home and starting a new job are associated with shifts away from the car-oriented modality style, although both effects are not very strong. This however does indicate that people start making different mode choices after they start working from home, which is a relevant finding given the large increase in working from home during and after the COVID-19 pandemic.\u003c/p\u003e \u003cp\u003eThere are relatively strong effects of both mobility tool ownership in wave 1 and changes in mobility tool ownership between the two waves on the transition probabilities. When interpreting these coefficients however, we must be careful regarding the assumed direction of causality. For example, we might interpret the negative coefficient (-1.08) of \u0026lsquo;always having access to car\u0026rsquo; in wave 1 on the transition from car-oriented to multi-modal modality styles in two distinct ways: first, that always having access to a car prevents people from making this transition. This assumes that mobility tool ownership is a cause of our travel behaviour and our changes therein. Second, that people who are otherwise disinclined to make such a transition, and thus probably are relatively car-dependent, are more likely to ensure they always have access to a car. This assumes that our travel behaviour, and especially our habitual patterns, causally affect our mobility tool ownership. In practice, both directions are likely to exist (Nurul Habib et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Scott \u0026amp; Axhausen, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). Below, we will try to keep both options in mind, but for reasons of legibility will not discuss both options for each coefficient.\u003c/p\u003e \u003cp\u003eThe first thing to notice is that all effects are in the expected direction, given the interpretation of the latent classes as modality styles: people who always have access to a car are less likely to transition from the car-oriented to the multi-modal modality style, whereas people who own electric bicycles or public transport cards are more likely to do so. Similarly, the effects of changes in mobility tool ownership also follow the expected direction. These effects are relatively strong as well: both gaining ownership of a personal car (2.51) and the closely related gaining the ability to always access a car (2.84) make it much more likely that someone transitions from the multi-modal to the car-oriented modality style. Gaining ownership of an electric bicycle makes people more likely to switch from car-oriented to multi-modal modality styles (1.76).\u003c/p\u003e \u003cp\u003eThese transition parameters can be used to calculate transition probabilities. Using these transition probabilities, we can create transition matrices, which are shown in Table\u0026nbsp;\u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e8\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eTransition matrices for respondents without life-events and for those with changes in mobility tool ownership\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" morerows=\"2\" nameend=\"c2\" namest=\"c1\" rowspan=\"3\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003ePeople \u003cem\u003ewithout\u003c/em\u003e life-events or changes in mobility tool ownership\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003ePeople who gained a car\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003eWave 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eWave 2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClass 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eClass 2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eClass 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eClass 2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eWave 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClass 1:\u003c/p\u003e \u003cp\u003eCar-\u003c/p\u003e \u003cp\u003eoriented\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.939\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0609\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eClass 1:\u003c/p\u003e \u003cp\u003eCar-oriented\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.0337\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClass 2:\u003c/p\u003e \u003cp\u003eMulti-modal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0723\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.928\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eClass 2:\u003c/p\u003e \u003cp\u003eMulti-modal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.264\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.736\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" morerows=\"2\" nameend=\"c2\" namest=\"c1\" rowspan=\"3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003ePeople who lost a car\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003ePeople who gained an electric bicycle\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003eWave 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eWave 2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClass 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eClass 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eClass 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eClass 2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWave 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClass 1:\u003c/p\u003e \u003cp\u003eCar-\u003c/p\u003e \u003cp\u003eoriented\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.153\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eClass 1:\u003c/p\u003e \u003cp\u003eCar-oriented\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.167\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClass 2:\u003c/p\u003e \u003cp\u003eMulti-modal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0337\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eClass 2:\u003c/p\u003e \u003cp\u003eMulti-modal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.00929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.991\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAs can be seen in the upper-left quadrant of Table \u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, the class membership of people who do not undergo any life-events or changes in mobility tool ownership is very stable, as roughly 94% and 92% of car-oriented and multi-modal people remain in their respective modality styles. This picture shifts dramatically for those who gain or lose a car and for those who gain an electric bicycle. Two things stand out: first, that gaining a car leads to a relatively higher chance to transition towards a car-oriented modality style (26%) than losing one leads to a transition towards a multi-modal modality style (15%). Car-oriented behaviour therefore seems to be more stable than multi-modal travel behaviour, and there is some asymmetry in the effect of vehicle ownership. Second, that gaining an electric bicycle is fairly effective at getting people to transition towards a multi-modal modality style.\u003c/p\u003e \u003cp\u003eA final result with respect to the transition probabilities is that we find relatively weak effects of life-events on transition probabilities. Perhaps these life-events don\u0026rsquo;t lead to changes in \u003cem\u003emode choice\u003c/em\u003e, but only to changes in \u003cem\u003emode use\u003c/em\u003e. This can be the result of changes in the activity-pattern generation, for example by affecting either trip generation or trip distribution rather than mode choice. Residential relocations to urbanized areas for example might lead individuals to make more shorter-distance trips, which are more likely to be made using the bicycle and walking. However, there need not be a change in sensitivity to travel distance and thus no large change in behavioural parameters. Given the same trip, the respondent would still make roughly the same choices. However, the types of trips made might have changed. Another explanation might be that life-events coincide with changes in vehicle-ownership. As we explicitly model the effects of vehicle-ownership, this indirect effect will not show up in the model. However, the corelations between life-events and changes in vehicle ownership were relatively small (\u0026lt;\u0026thinsp;0.2). Therefore this explanation is unlikely to fully explain the weak effects of life-events.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e4.5. Enumeration of transition effects on choice probabilities\u003c/h2\u003e \u003cp\u003eTo get a more intuitive understanding of the meaning of the transitions between the modality styles, we used the estimated LC transition model to calculate mode choice probabilities for individuals with varying states of vehicle ownership. As a result, we can see the effect that changes in vehicle ownership have on the mode choice probabilities. We illustrate this result using two different approaches. First, to show the effect of changes in car access, we show the estimated choice probabilities for trips made by people grouped by car ownership. As people\u0026rsquo;s car ownership changes, so does their estimated probability of belonging to a certain modality style. Due to these transitions between the modality styles, their estimated choice probabilities change as well as is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eA couple of observations can be made based on Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. First, car ownership is a large determinant of travel behaviour: people who own a car are much more likely to choose the car. Second, the effect of buying a car is much larger than the effect of selling a car. The increase in car probability is much larger for people who bought a car than the decrease for people who sold a car. Third, there are substantial lead effects: people who buy a car between the two waves already use the car much more often than people who do not do so.\u003c/p\u003e \u003cp\u003eFor the second illustration, we again use the reference trips within each kilometre band, which we used earlier to illustrate the difference between the modality styles. Now, we plot the unconditional bicycle choice probability of people grouped by e-bike ownership. This allows us to show how the probability of choosing the bicycle across various distance ranges changes as a result of differences in e-bike ownership.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe want to highlight two results here. First, that buying an electric bicycle increases the probability of choosing the bicycle. For the two groups whose electric bicycle ownership did not change, we observe no change in mode choice probability for the bicycle. For the group that did buy an e-bike, we observed a sizeable increase in the predicted market share of the bicycle from the first to the second wave. Model estimates of the treatment effect of the untreated, that is, people who did not already buy an electric bike, of buying an electric bicycle are roughly in the range of 8 percentage points, with the mode share of the bicycle increasing from 33\u0026ndash;41%. This increase corresponds with a decrease in the estimated mode share of the car. As such, our results indicate that the electric bicycle substitutes car use, especially for shorter-distance trips up to roughly 15km in length.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eIn this paper, we used a latent transition choice model to estimate the longitudinal transitions between modality styles. Two latent classes are found, which can be identified as two distinct modality styles, namely \u0026lsquo;Car-oriented\u0026rsquo; and \u0026lsquo;Multimodal\u0026rsquo;. The \u0026lsquo;car-oriented\u0026rsquo; modality style is found to be more sensitive to travel time increases than the multimodal modality style. The modality styles are relatively stable over time, especially in the absence of any life-events or changes in modality styles. The car-oriented modality style is found to be more stable than the multi-modal modality style. Life-events only have relatively minor effects on the transition probabilities between the modality styles: people who started a new job or increased the hours they worked from home moved from the car-oriented to the multimodal class slightly more often. These findings seem to contradict earlier studies showing relatively larger effects of life-events on travel behaviour (Clark et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Gao et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Olde Kalter et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Rau \u0026amp; Manton, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). These previous studies typically used clustering approaches, which cannot disentangle trip generation changes from mode choice changes. Our results therefore suggest that life events mostly affect trip generation rather than the mode choice process itself.\u003c/p\u003e \u003cp\u003eOwnership of mobility tools and changes therein have much larger effects on the transition probabilities. Whilst the existence and direction of these effects are not wholly surprising (car ownership increases the probability one belongs to the car-oriented modality style, and electric bicycle ownership increases the probability one belongs to the multi-modal modality style), some findings are still very noteworthy.\u003c/p\u003e \u003cp\u003eFirst, we find clear evidence for an asymmetry in the effect of car ownership, where gaining a car has a much larger effect on the transition probabilities than losing a car. As a result, attaining ownership of a car seems to have an irreversible effect on one\u0026rsquo;s travel behaviour, even if the car has to be sold. From a policy perspective \u0026ndash; assuming that reducing overall car use is the policy objective \u0026ndash; it is therefore important to facilitate lifestyles that do not depend on car ownership and provide car-less people with similar levels of accessibility as car owners. Second, we find evidence of either lead- or self-selection effects, where people who buy a car or bicycle in-between the two waves already respectively used the car or the bicycle more often in the first wave than those who did not buy such a vehicle. These findings point to the importance of establishing good counterfactuals when studying the effects of vehicle ownership on travel behaviour, for example by using longitudinal data. Third, we find that, even after controlling for these selection effects, buying an electric bicycle results in a noticeable shift towards the more multi-modal modality style. The transition towards the less time- and distance sensitive multi-modal modality style also suggests that buying an electric bicycle makes bicyclists less sensitive to increases in travel time and travel distance. As a result, the bicycle choice probability increases substantially, especially for shorter-distance trips up to roughly 15km in length. These findings complement earlier studies using longitudinal clustering and structural equation modeling methods (de Haas et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Kroesen, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe latent transition choice model thus enables us to improve further our understanding of the effects of changes in mobility tool ownership on mode choice behaviour. However, even though the model captures dynamic effects and thus uses the possibilities of panel data, the causal direction is still difficult to establish. This is due to the yearly occurrence of the data, which means that changes in mobility tool ownership and changes in choice behaviour seem to coincide together. This ambiguity makes it more difficult to draw clear behavioural conclusions. A second drawback of the current approach relates to the relatively low occurrence of life-events and changes in mobility tool ownership. As a result, the power of the model to reliably assess the effects of these changes on mode choice behaviour is limited. The estimation of the model is also made more difficult, as it is dependent on relatively limited observations where changes in choice behaviour occur. A final limitation we would like to highlight is that the current model is only estimable with two latent classes. As the number of latent classes increases, the number of transitions between classes increases exponentially. To illustrate, even a three-class solution would require the estimation of six transition parameters. Combined with the difficulties mentioned above, this was not feasible. The downside of a two-class solution is that they might oversimplify the existing heterogeneity with respect to mode choice behaviour.\u003c/p\u003e \u003cp\u003eThere are several areas for future research that seem worthwhile. First, we could estimate a model using more longitudinal waves. This could prove especially fruitful given the relatively low occurrence of life-events. The ability to use more data could enable us to provide more reliable estimates of the effects of changes in life-events on travel behaviour. Second, we could add more detailed life-events and perhaps model changes of the transport system as well. Examples of the first type could include whether a residential relocation moved towards a more car-oriented or a more multi-modal oriented residential environment and a related example of the second type is to study whether autonomous changes in the built environment have an impact on the transition probabilities. More generally, longitudinal choice models can be used to estimate the stability of preferences regarding attributes such as time and cost, which could be interesting to empirically estimate longitudinal effects of, for example, changes in income on the value of travel time. Finally, use of more intensive longitudinal data could allow for a better understanding of the directions of causality involved. The yearly waves in the MPN are unable to capture when exactly behavioural shifts occurred, and if they preceded or followed life-events or changes in mobility tool ownership. Daily longitudinal data, for example using GPS tracking devices, combined with more complex travel surveys could help to further our understanding of the causality involved.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThe authors received no specific funding for this work\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eRF - Conceptualization, Methodology, Software, Formal Analysis, Data Curation, Writing - original draft, VisualizationSvC - Methodology, Software, Writing - Review \u0026amp; Editing, SupervisionMK - Conceptualization, Writing - Review \u0026amp; Editing, SupervisionEM - Conceptualization, Writing - Review \u0026amp; Editing, Supervision\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAccess to data from the Netherlands Mobility Panel can be requested through https://mpndata.nl/. The specific data file used to support the analyses in this paper, as well as the scripts necessary to reproduce it from the base data files in the Netherlands Mobility Panel, are saved by the corresponding author.\u003c/p\u003e\u003ch2\u003eCompeting Interests\u003c/h2\u003e\n\u003cp\u003eThe authors have no conflicts of interest, financial or otherwise, to declare that are relevant to the content of this article.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAarts, H., Dijksterhuis, A.: The automatic activation of goal-directed behaviour: the case of travel habit. J. Environ. Psychol. \u003cb\u003e20\u003c/b\u003e(1), 75\u0026ndash;82 (2000). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1006/JEVP.1999.0156\u003c/span\u003e\u003cspan address=\"10.1006/JEVP.1999.0156\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnderson, T.W.: Probability Models for Analyzing Time Changes In Attitudes. Free Press. (1954). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://scholar.google.nl/scholar?hl=nl\u0026amp;as_sdt=0%2C5\u0026amp;q=Probability+Models+for+Analyzing+Time+Changes+In+Attitudes\u0026amp;btnG\u003c/span\u003e\u003cspan address=\"https://scholar.google.nl/scholar?hl=nl\u0026amp;as_sdt=0%2C5\u0026amp;q=Probability+Models+for+Analyzing+Time+Changes+In+Attitudes\u0026amp;btnG\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e=\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBen-Akiva, M., Lerman, S.R.: Discrete choice analysis: theory and application to travel demand, 1st edn. MIT Press (1985). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://trid.trb.org/view/274564\u003c/span\u003e\u003cspan address=\"https://trid.trb.org/view/274564\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBen-Akiva, M., Mcfadden, D., Abe, M., B\u0026ouml;ckenholt, U., Bolduc, D., Revelt, D., Steinberg, D.: Modeling Methods for Discrete Choice Analysis. Mark. Lett. \u003cb\u003e8\u003c/b\u003e, 273\u0026ndash;286 (1997)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBhat, C.R.: Incorporating Observed and Unobserved Heterogeneity in Urban Work Travel Mode Choice Modeling. \u003cem\u003eHttps://Doi.Org\u003c/em\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e/10.1287/Trsc.34.2.228\u003c/span\u003e\u003cspan address=\"http:///10.1287/Trsc.34.2.228\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003cem\u003e12306\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(2), 228\u0026ndash;238. (2000). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1287/TRSC.34.2.228.12306\u003c/span\u003e\u003cspan address=\"10.1287/TRSC.34.2.228.12306\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eB\u0026ouml;ckenholt, U., Langeheine, R.: Latent change in recurrent choice data. Psychometrika. \u003cb\u003e61\u003c/b\u003e(2), 285\u0026ndash;301 (1996). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/BF02294340/METRICS\u003c/span\u003e\u003cspan address=\"10.1007/BF02294340/METRICS\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarrasco, J.A., De Dios Ort\u0026uacute;zar, J., De, J., Ortu, D., Zar, \u0026Acirc;.: Review and assessment of the nested logit model. Transp. Reviews. \u003cb\u003e22\u003c/b\u003e(2), 197\u0026ndash;218 (2002). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01441640110091224\u003c/span\u003e\u003cspan address=\"10.1080/01441640110091224\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCherchi, E., Cirillo, C., de Ort\u0026uacute;zar, J. D: Modelling correlation patterns in mode choice models estimated on multiday travel data. Transp. Res. Part. A: Policy Pract. \u003cb\u003e96\u003c/b\u003e, 146\u0026ndash;153 (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRA.2016.11.021\u003c/span\u003e\u003cspan address=\"10.1016/J.TRA.2016.11.021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCirillo, C., Axhausen, K.W.: Evidence on the distribution of values of travel time savings from a six-week diary. Transp. Res. Part. A: Policy Pract. \u003cb\u003e40\u003c/b\u003e(5), 444\u0026ndash;457 (2006). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRA.2005.06.007\u003c/span\u003e\u003cspan address=\"10.1016/J.TRA.2005.06.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClark, B., Chatterjee, K., Melia, S., Knies, G., Laurie, H.: Life events and travel behavior exploring the interrelationship using UK Household Longitudinal Study data. Transp. Res. Rec. \u003cb\u003e2413\u003c/b\u003e, 54\u0026ndash;64 (2014). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3141/2413-06\u003c/span\u003e\u003cspan address=\"10.3141/2413-06\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaly, A., Zachary, S.: Improved multiple choice models. In D. Hensher \u0026amp; O. Dalvi (Eds.), Identifying and Measuring the Determinants of Model Choice (pp. 187\u0026ndash;201). (1979). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.researchgate.net/publication/230663926\u003c/span\u003e\u003cspan address=\"https://www.researchgate.net/publication/230663926\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Haas, M.C., Kroesen, M., Chorus, C., Hoogendoorn-Lanser, S., Hoogendoorn, S.: E-bike user groups and substitution effects: evidence from longitudinal travel data in the Netherlands. Transportation. \u003cb\u003e49\u003c/b\u003e(3), 815\u0026ndash;840 (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/S11116-021-10195-3/FIGURES/3\u003c/span\u003e\u003cspan address=\"10.1007/S11116-021-10195-3/FIGURES/3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Haas, M.C., Scheepers, C.E., Harms, L.W.J., Kroesen, M.: Travel pattern transitions: Applying latent transition analysis within the mobility biographies framework. Transp. Res. Part. A: Policy Pract. \u003cb\u003e107\u003c/b\u003e, 140\u0026ndash;151 (2018). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tra.2017.11.007\u003c/span\u003e\u003cspan address=\"10.1016/j.tra.2017.11.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiana, M., Mokhtarian, P.L.: Grouping travelers on the basis of their different car and transit levels of use. Transportation. \u003cb\u003e36\u003c/b\u003e(4), 455\u0026ndash;467 (2009). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11116-009-9207-y\u003c/span\u003e\u003cspan address=\"10.1007/s11116-009-9207-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFaber, R.M., Jonkeren, O., de Haas, M.C., Molin, E.J.E., Kroesen, M.: Inferring modality styles by revealing mode choice heterogeneity in response to weather conditions. Transp. Res. Part. A: Policy Pract. \u003cb\u003e162\u003c/b\u003e, 282\u0026ndash;295 (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRA.2022.06.003\u003c/span\u003e\u003cspan address=\"10.1016/J.TRA.2022.06.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFosgerau, M.: Investigating the distribution of the value of travel time savings. Transp. Res. Part. B: Methodological. \u003cb\u003e40\u003c/b\u003e(8), 688\u0026ndash;707 (2006). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRB.2005.09.007\u003c/span\u003e\u003cspan address=\"10.1016/J.TRB.2005.09.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao, J., He, S.Y., Ettema, D., Helbich, M.: Travel behavior changes due to life events: Longitudinal evidence from Dutch couple households. Transp. Res. Part. A: Policy Pract. \u003cb\u003e175\u003c/b\u003e, 103765 (2023). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRA.2023.103765\u003c/span\u003e\u003cspan address=\"10.1016/J.TRA.2023.103765\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eG\u0026auml;rling, T., Axhausen, K.W.: Introduction: Habitual travel choice. Transportation. \u003cb\u003e30\u003c/b\u003e(1), 1\u0026ndash;11 (2003). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1023/A:1021230223001/METRICS\u003c/span\u003e\u003cspan address=\"10.1023/A:1021230223001/METRICS\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHamaker, E.L.: Why researchers should think within-person: A paradigmatic rationale RI-CLPM. In Handbook of Research Methods for Studying Daily life (pp. 43\u0026ndash;61). (2012). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.researchgate.net/publication/266896375\u003c/span\u003e\u003cspan address=\"https://www.researchgate.net/publication/266896375\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHensher, D.A., Greene, W.H.: The mixed logit model: The state of practice. Transportation. \u003cb\u003e30\u003c/b\u003e(2), 133\u0026ndash;176 (2003). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1023/A:1022558715350/METRICS\u003c/span\u003e\u003cspan address=\"10.1023/A:1022558715350/METRICS\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHess, S., Daly, A.J.: Handbook of choice modelling. Edw. Elgar publishing (2014). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.e-elgar.com/shop/gbp/handbook-of-choice-modelling-9781781003145.html\u003c/span\u003e\u003cspan address=\"https://www.e-elgar.com/shop/gbp/handbook-of-choice-modelling-9781781003145.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHess, S., Palma, D.: Apollo: a flexible, powerful and customisable freeware package for choice model estimation and application version 0.0.7 User manual. www.ApolloChoiceModelling.com (2019)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHess, S., Palma, D., Calastri, C., Crasted dit Sourd, R., Daly, A., Dumont, J., Molloy, J., Schmid, B.: Apollo: a flexible, powerful and customisable freeware package for choice model estimation and application. In Journal of Choice Modelling: Vol. In Press. Elsevier. (2019). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.JOCM.2019.100170\u003c/span\u003e\u003cspan address=\"10.1016/J.JOCM.2019.100170\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoogendoorn-Lanser, S., Schaap, N.T.W., Olde Kalter, M.J.: The netherlands mobility panel: An innovative design approach for web-based longitudinal travel data collection. Transp. Res. Procedia. \u003cb\u003e11\u003c/b\u003e, 311\u0026ndash;329 (2015). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.trpro.2015.12.027\u003c/span\u003e\u003cspan address=\"10.1016/j.trpro.2015.12.027\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJanke, J., Handy, S.: How life course events trigger changes in bicycling attitudes and behavior: Insights into causality. Travel Behav. Soc. \u003cb\u003e16\u003c/b\u003e, 31\u0026ndash;41 (2019). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TBS.2019.03.004\u003c/span\u003e\u003cspan address=\"10.1016/J.TBS.2019.03.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeskisaari, V., Ottelin, J., Heinonen, J.: Greenhouse gas impacts of different modality style classes using latent class travel behavior model. J. Transp. Geogr. \u003cb\u003e65\u003c/b\u003e, 155\u0026ndash;164 (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jtrangeo.2017.10.018\u003c/span\u003e\u003cspan address=\"10.1016/j.jtrangeo.2017.10.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKl\u0026oslash;ckner, C.: How single events change travel mode choice: a life span perspective. (2013). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/2392913\u003c/span\u003e\u003cspan address=\"https://ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/2392913\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKroesen, M.: Modeling the behavioral determinants of travel behavior: An application of latent transition analysis. Transp. Res. Part. A: Policy Pract. \u003cb\u003e65\u003c/b\u003e, 56\u0026ndash;67 (2014). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tra.2014.04.010\u003c/span\u003e\u003cspan address=\"10.1016/j.tra.2014.04.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKroesen, M.: To what extent do e-bikes substitute travel by other modes? Evidence from the Netherlands. Transp. Res. Part. D: Transp. Environ. \u003cb\u003e53\u003c/b\u003e, 377\u0026ndash;387 (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.trd.2017.04.036\u003c/span\u003e\u003cspan address=\"10.1016/j.trd.2017.04.036\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiao, F., Molin, E., Timmermans, H., van Wee, B.: The impact of business models on electric vehicle adoption: A latent transition analysis approach. Transp. Res. Part. A: Policy Pract. \u003cb\u003e116\u003c/b\u003e, 531\u0026ndash;546 (2018). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRA.2018.07.008\u003c/span\u003e\u003cspan address=\"10.1016/J.TRA.2018.07.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoder, A., Axhausen, K.W.: Mobility tools and use: Accessibility\u0026rsquo;s role in Switzerland. J. Transp. Land. Use. \u003cb\u003e11\u003c/b\u003e(1), 367\u0026ndash;385 (2018). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5198/JTLU.2018.1054\u003c/span\u003e\u003cspan address=\"10.5198/JTLU.2018.1054\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLouviere, J., Hensher, D.A., Swait, J.D.: Stated choice methods: analysis and application. Cambridge University Press (2000). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/CBO9780511753831.008\u003c/span\u003e\u003cspan address=\"10.1017/CBO9780511753831.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcFadden, D.: Conditional logit analysis of qualitative choice behavior. Front. Econometrics, 105\u0026ndash;142. (1974). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://elsa.berkeley.edu/reprints/mcfadden/zarembka.pdf\u003c/span\u003e\u003cspan address=\"http://elsa.berkeley.edu/reprints/mcfadden/zarembka.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcNally, M.G.: The Four Step Model, Handbook of Transport Modeling. In Institute of Transportation Studies and Department of Civil \u0026amp; Environmental Engineering. (2000). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://escholarship.org/uc/item/7j0003j0\u003c/span\u003e\u003cspan address=\"https://escholarship.org/uc/item/7j0003j0\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMolin, E., Mokhtarian, P., Kroesen, M.: Multimodal travel groups and attitudes: A latent class cluster analysis of Dutch travelers. Transp. Res. Part. A: Policy Pract. \u003cb\u003e83\u003c/b\u003e, 14\u0026ndash;29 (2016). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tra.2015.11.001\u003c/span\u003e\u003cspan address=\"10.1016/j.tra.2015.11.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eM\u0026uuml;ggenburg, H., Busch-Geertsema, A., Lanzendorf, M.: Mobility biographies: A review of achievements and challenges of the mobility biographies approach and a framework for further research. J. Transp. Geogr. \u003cb\u003e46\u003c/b\u003e, 151\u0026ndash;163 (2015). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.JTRANGEO.2015.06.004\u003c/span\u003e\u003cspan address=\"10.1016/J.JTRANGEO.2015.06.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMurphy, J.J., Allen, P.G., Stevens, T.H., Weatherhead, D.: A meta-analysis of hypothetical bias in stated preference valuation. Environ. Resource Econ. \u003cb\u003e30\u003c/b\u003e(3), 313\u0026ndash;325 (2005). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/S10640-004-3332-Z/METRICS\u003c/span\u003e\u003cspan address=\"10.1007/S10640-004-3332-Z/METRICS\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeal, D.T., Wood, W., Labrecque, J.S., Lally, P.: How do habits guide behavior? Perceived and actual triggers of habits in daily life. J. Exp. Soc. Psychol. \u003cb\u003e48\u003c/b\u003e(2), 492\u0026ndash;498 (2012). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.JESP.2011.10.011\u003c/span\u003e\u003cspan address=\"10.1016/J.JESP.2011.10.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNurul Habib, K., Weiss, A., Hasnine, S.: On the heterogeneity and substitution patterns in mobility tool ownership choices of post-secondary students: The case of Toronto. (2018). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tra.2018.06.002\u003c/span\u003e\u003cspan address=\"10.1016/j.tra.2018.06.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOlde Kalter, M.J., Paix Puello, L., L., Geurs, K.T.: Exploring the relationship between life events, mode preferences and mode use of young adults: A 3-year cross-lagged panel analysis in the Netherlands. Travel Behav. Soc. \u003cb\u003e24\u003c/b\u003e, 195\u0026ndash;204 (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tbs.2021.04.004\u003c/span\u003e\u003cspan address=\"10.1016/j.tbs.2021.04.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrato, C.G., Halld\u0026oacute;rsd\u0026oacute;ttir, K., Nielsen, O.A.: Latent lifestyle and mode choice decisions when travelling short distances. Transportation. \u003cb\u003e44\u003c/b\u003e(6), 1343\u0026ndash;1363 (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11116-016-9703-9\u003c/span\u003e\u003cspan address=\"10.1007/s11116-016-9703-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRau, H., Manton, R.: Life events and mobility milestones: Advances in mobility biography theory and research. (2016). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jtrangeo.2016.02.010\u003c/span\u003e\u003cspan address=\"10.1016/j.jtrangeo.2016.02.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScheiner, J.: Mobility Biographies and Mobility Socialisation\u0026mdash;New Approaches to an Old Research Field. Life-Oriented Behavioral Research for Urban Policy, 385\u0026ndash;401. (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-4-431-56472-0_13\u003c/span\u003e\u003cspan address=\"10.1007/978-4-431-56472-0_13\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchwanen, T., Banister, D., Anable, J.: Rethinking habits and their role in behaviour change: the case of low-carbon mobility. J. Transp. Geogr. \u003cb\u003e24\u003c/b\u003e, 522\u0026ndash;532 (2012). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.JTRANGEO.2012.06.003\u003c/span\u003e\u003cspan address=\"10.1016/J.JTRANGEO.2012.06.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScott, D.M., Axhausen, K.W.: Household mobility tool ownership: Modeling interactions between cars and season tickets. Transportation. \u003cb\u003e33\u003c/b\u003e(4), 311\u0026ndash;328 (2006). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/S11116-005-0328-7/METRICS\u003c/span\u003e\u003cspan address=\"10.1007/S11116-005-0328-7/METRICS\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmall, K.A.: Valuation of travel time. Econ. Transp. \u003cb\u003e1\u003c/b\u003e(1\u0026ndash;2), 2\u0026ndash;14 (2012). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.ECOTRA.2012.09.002\u003c/span\u003e\u003cspan address=\"10.1016/J.ECOTRA.2012.09.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStr\u0026ouml;mberg, H., Rexfelt, O., Karlsson, I.C.M.A., Sochor, J.: Trying on change \u0026ndash; Trialability as a change moderator for sustainable travel behaviour. Travel Behav. Soc. \u003cb\u003e4\u003c/b\u003e, 60\u0026ndash;68 (2016). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TBS.2016.01.002\u003c/span\u003e\u003cspan address=\"10.1016/J.TBS.2016.01.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrain, K.E.: Discrete choice methods with simulation, second edition. In Discrete Choice Methods with Simulation, Second Edition (Vol. 9780521766). (2009). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/CBO9780511805271\u003c/span\u003e\u003cspan address=\"10.1017/CBO9780511805271\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Cranenburgh, S., Kouwenhoven, M.: An artificial neural network based method to uncover the value-of-travel-time distribution. Transportation. \u003cb\u003e48\u003c/b\u003e(5), 2545\u0026ndash;2583 (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/S11116-020-10139-3/TABLES/9\u003c/span\u003e\u003cspan address=\"10.1007/S11116-020-10139-3/TABLES/9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVij, A., Carrel, A., Walker, J.L.: Incorporating the influence of latent modal preferences on travel mode choice behavior. Transp. Res. Part. A: Policy Pract. \u003cb\u003e54\u003c/b\u003e, 164\u0026ndash;178 (2013). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tra.2013.07.008\u003c/span\u003e\u003cspan address=\"10.1016/j.tra.2013.07.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWardman, M., Chintakayala, V.P.K., de Jong, G.: Values of travel time in Europe: Review and meta-analysis. Transp. Res. Part. A: Policy Pract. \u003cb\u003e94\u003c/b\u003e, 93\u0026ndash;111 (2016). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/J.TRA.2016.08.019\u003c/span\u003e\u003cspan address=\"10.1016/J.TRA.2016.08.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiggins, L.M.: Mathematical models for the interpretation of attitude and behavior change: The analysis of multi-wave panel. Columbia University ProQuest Dissertations \u0026amp; Theses. (1955)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliams, H.C.W.L.: On the Formation of Travel Demand Models and Economic Evaluation Measures of User Benefit. (1977). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eHttp://Dx.Doi.Org/10.1068/A090285\u003c/span\u003e\u003cspan address=\"Http://Dx.Doi.Org/10.1068/A090285\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, 9(3), 285\u0026ndash;344. https://doi.org/10.1068/A090285\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiong, C., Chen, X., He, X., Guo, W., Zhang, L.: The analysis of dynamic travel mode choice: a heterogeneous hidden Markov approach. Transportation. \u003cb\u003e42\u003c/b\u003e(6), 985\u0026ndash;1002 (2015). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/S11116-015-9658-2/TABLES/6\u003c/span\u003e\u003cspan address=\"10.1007/S11116-015-9658-2/TABLES/6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZarwi, E., Vij, W., Zarwi, F., El, Vij, A., Walker, J.L.: Modeling and Forecasting the Evolution of Preferences over Time. A Hidden Markov Model of Travel Behavior (2017)\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Behavioural change, Discrete Choice Modelling, Modality Styles, Mobility Biographies, Netherlands Mobility Panel","lastPublishedDoi":"10.21203/rs.3.rs-5353959/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5353959/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMode choice is an essential subject within travel behaviour research. Typically, mode choice has been studied using cross-sectional (discrete choice) models, which assume that all choices are made simultaneously. In this study, we relax this assumption by explicitly incorporating the time when a choice is made within the modeling framework, using a latent transition choice model. This model allows for the identification of the effects of life-events and (changes in) mobility tool ownership on mode choice probabilities over time. To estimate the model, data from the Netherlands Mobility Panel gathered between 2016 and 2022 are used. The model identifies two latent classes, 1) a car-dependent modality style and 2) a multi-modal modality style. The transition probabilities between these classes in-between two consecutive waves are estimated, as well as the effects of life-events and mobility tool ownership on these transition probabilities. We find substantial and statistically significant effects from changes in vehicle ownership on the transition probabilities, indicating that electric bicycle ownership leads to a substitution of the car towards the bicycle on shorter-distance trips even after controlling for lead- and self-selection effects.\u003c/p\u003e","manuscriptTitle":"The effects of life-events and changes in mobility tool ownership on mode choice behaviour","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-11-11 13:50:05","doi":"10.21203/rs.3.rs-5353959/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8c640560-ad3e-4774-ad68-0b4787370b8f","owner":[],"postedDate":"November 11th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-08-14T17:53:23+00:00","versionOfRecord":[],"versionCreatedAt":"2024-11-11 13:50:05","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5353959","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5353959","identity":"rs-5353959","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}