Content uploaded by Otto Anker Nielsen
Author content
All content in this area was uploaded by Otto Anker Nielsen on Dec 06, 2017
Content may be subject to copyright.
Latent lifestyle and mode choice decisions
when travelling short distances
Carlo Giacomo Prato
1
•Katrı
´n Halldo
´rsdo
´ttir
2
•
Otto Anker Nielsen
2
ÓSpringer Science+Business Media New York 2016
Abstract In the quest for sustainable travel, short distances appear the most amenable to
curbing the use of the automobile. Existing studies about short trips evaluate the potential
of shifting from the automobile to sustainable travel options while considering the popu-
lation as homogeneous in its preferences and its tendency to accept these alternative travel
options as realistic. However, this assumption appears quite unrealistic and the current
study offers a different perspective: the mode choices when travelling short distances are
likely related to lifestyle decisions. Short trip chains of a representative sample of the
Danish population in the Copenhagen Region were analysed, and more specifically a latent
class choice model was estimated to uncover latent lifestyle groups and choice specific
travel behaviour. Results show that four lifestyle groups are identified in the population:
car oriented, bicycle oriented, public transport oriented and public transport averse. Each
lifestyle group has specific perceptions of travel time (with extremely different rates of
substitution between alternative travel modes), transfer penalties in public transport trip
chains, weather influence (especially on active travel modes), and trip purpose effect on
mode selection. Consequently, when thinking about measures to increase the appeal of
sustainable travel options, decision-makers should look at specific individuals within the
population and more sensitive individuals to comfort and level-of-service improvements
across the lifestyle groups.
Keywords Lifestyle choices Short distances Mode choice Latent class models Car
oriented Bicycle oriented Public transport oriented
A preliminary version of the study was presented at the IATBR conference in Windsor in July 2015, and an
earlier version of the manuscript was significantly improved thanks to the insightful comments of three
anonymous reviewers.
&Carlo Giacomo Prato
c.prato@uq.edu.au
1
School of Civil Engineering, The University of Queensland, Brisbane, QLD 4072, Australia
2
Department of Transport, Technical University of Denmark, Bygningstorvet 116 B,
2800 Kgs. Lyngby, Denmark
123
Transportation
DOI 10.1007/s11116-016-9703-9
Introduction
When looking at mode choice decisions, the core assumption is that individuals want to
travel from their origin to their destination in the way that guarantees the shortest time, the
cheapest cost, the most comfortable travel, and the most flexible opportunity for escorting
children to their activities, carrying heavy goods and changing destination or time of travel
seamlessly. The automobile provides the answers to these needs, as its large use in cities
throughout the world has been observed even for short trips where it has replaced sus-
tainable travel options such as walking, cycling, and public transport (see, e.g., Pucher
et al. 1999; Rietveld 2000; Mackett 2003; de Nazelle et al. 2010).
The large use of the automobile causes great distress for the sustainability of the cities
of the future, given that environmental (see, e.g., Hertel et al. 2008; de Nazelle et al. 2010;
Lindsay et al. 2011), climate (see, e.g., Maibach et al. 2009; Fuglestvedt et al. 2010;
Borken-Kleefeld et al. 2013), and health (see, e.g., de Nazelle et al. 2011; Rojas-Rueda
et al. 2011; Scheepers et al. 2013) concerns urge looking for sustainable travel solutions.
While the convenience and swiftness of the automobile might thwart the attempt to reduce
its use in suburban or rural areas where activities are dispersed over larger distances and
travel alternatives are more scarce, short trips appear more amenable to curbing automobile
use (e.g., Frank et al. 2000; Mackett et al. 2003; Loukopoulos and Ga
¨rling 2005; Kim and
Ulfarsson 2008; de Nazelle et al. 2010; Monzon et al. 2011).
Modal shift from the automobile to sustainable travel modes for short distances has been
analysed in the literature with a focus on the potential for individuals to benefit from the
climate, environmental, and health perspective. However, this potential very rarely
translates into individuals actually leaving their cars, as the automobile is very convenient
even on short trips for carrying goods, picking up and dropping off spouses and children,
staying within time constraints, and enjoying comfort and convenience. Even though
emission reduction is an obvious positive effect of the use of public transport and active
travel modes (e.g., Frank et al. 2000; de Nazelle et al. 2010; Lindsay et al. 2011), often
individuals value the convenience and swiftness of the automobile with respect to these
sustainable travel options far more than their potential contribution to solving environ-
mental issues (e.g., Banister 2008; Monzon et al. 2011; Borken-Kleefeld et al. 2013). Even
though active travel present obvious benefits from the health perspective (e.g., Rojas-
Rueda et al. 2011; Grabow et al. 2012; Piatkowski et al. 2015), often individuals consider
the potential risks in terms of decreased safety and increased accident probability (e.g., de
Hartog et al. 2010; Rojas-Rueda et al. 2011; Schepers and Heinen 2013).
The assessment of the potential for modal shift from automobile use to active travel and/
or public transport for short trips relies usually on mode frequency analysis and mode
choice models (see, e.g., Mackett 2003; Loukopoulos and Ga
¨rling 2005; Kim and
Ulfarsson 2008; Maibach et al. 2009; de Nazelle et al. 2010; Rojas-Rueda et al. 2011;
Carse et al. 2013; Scheepers et al. 2013). Interestingly, individuals are assumed in these
studies to have the same taste preferences and hence the same probability of shifting travel
mode. The assumption of taste preference homogeneity does not seem plausible: (i) travel
behaviour has extensive literature on taste heterogeneity across individuals in mode choice
models (for recent applications see, e.g., Hess and Stathopoulos 2013; Forsey et al. 2014;
Noland et al. 2014; Paulssen et al. 2014); (ii) long-term decisions such as residential
location, workplace location, car ownership, bicycle possession and public transport pass
purchase play a role on short-term decisions such as mode choice (for recent applications
see, e.g., Pinjari et al. 2011; Vovsha et al. 2013; Zhou 2014; Guerra 2015). The assumption
Transportation
123
of all individuals having the same probability of shifting travel mode does not seem
plausible either, since their lifestyle most likely plays a role in the mode choice for short
distances just as it is observed to play a part in other decisions: (i) residential location has
been related to the lifestyle of households (e.g., Walker and Li 2007; Smith and Olaru
2013) and knowledge-workers (e.g., Frenkel et al. 2013a,b); (ii) mode choices and
mobility styles have been associated with the lifestyle of individuals and their correlated
residential locations (e.g., Krizek 2006; Scheiner and Holz-Rau 2007; Kitamura 2009; Vij
et al. 2013); (iii) the decision about owning a car and, in that case, selecting a car type has
been connected with lifestyle stages (e.g., Choo and Mokhtarian 2004; van Acker et al.
2014); (iv) time use patterns, activity participation and neighbourhood characteristics have
been linked to lifestyle choices (e.g., Krizek and Waddell 2002; Schwanen and Mokhtarian
2005; Fan and Khattak 2012; Sun et al. 2012); (v) risky driving in adolescents has been
coupled with the lifestyle of the family where they were raised (Bina et al. 2006).
The current study proposes the analysis of mode choices for short distances from a
lifestyle perspective. Unlike existing studies about short trips, this study recognises the
heterogeneity across individuals and relates the short-term choices of travel mode to the
long-term lifestyle decisions. Unlike most existing studies about lifestyle in the trans-
portation literature, this study looks at lifestyle influencing short-term choices rather than
long-term ones such as residential location or car availability, and analyses lifestyle with a
representative sample of the population rather than biased convenience samples that hardly
represent the areas where the studies are conducted. The current study proposes a latent
class analysis that allows inferring simultaneously how lifestyle affects the decision of how
to travel in a short-time horizon and for short distances where sustainable travel options
seem realistically feasible.
This study focuses on short trip chains in the Copenhagen region as an example of a
large metropolitan area that offers sustainable travel options and yet experiences significant
traffic congestion and stalling cycling modal shares. Although policies exist to curb car
purchase with extremely high registration taxes that have negative externalities (see, e.g.,
Mabit and Fosgerau 2011; Rich et al. 2013) and extensive infrastructure exists for cycling,
not only an impasse in the cycling shares has been observed, but also young Danes have
expressed the worrisome intention of using the car instead of the bicycle in the near future
(Sigurdardottir et al. 2013). The current study offers a different perspective when looking
at the impasse in sustainable travel progress by estimating a latent class model that links
observable characteristics of the individual to the likelihood of having chosen a certain
lifestyle that then affects the travel choices for short distances. Data about short trip chains
were available from a representative sample of the Danish population who participated in
the Danish National Travel Survey: the sample included 10,982 trip chains with five
available alternative modes (i.e., walking, cycling, car driver, car passenger, and public
transport), and contained information about the characteristics of the travellers, the trip
chains, and the environment.
The remainder of the paper is structured as follows. ‘‘Lifestyle definition and mea-
surement’’ section presents the behavioural hypothesis and the definition of lifestyle.
‘‘ Methods’’ section introduces the model formulation and the data collection for looking at
the effect of lifestyle on the mode choice of Copenhageners travelling short distances.
‘‘ Estimation results’’ section presents the empirical results in terms of determinants of
travel behaviour specific to individuals having different lifestyles and predictors of lifestyle
group belonging. ‘‘Summary and conclusions’’ section summarises the conclusions and
highlights further research directions.
Transportation
123
Lifestyle definition and measurement
The first explicit reference to lifestyle in travel behaviour research is credited to Salomon
(1980), which later proposed a definition of lifestyle as ‘‘A pattern of behavior which
conforms to the individual’s orientation toward the three major roles of: household
member, a worker, and a consumer of leisure, and which conforms to the constrained
resources available’’ (Salomon and Ben-Akiva 1983). However, there is not a formally
agreed definition of lifestyle because its elaboration is pragmatical rather than theoretical
(van Acker 2015).
When looking at lifestyle from a conceptual perspective, the current study looks at
lifestyle expressions as patterns of behaviour through which the individual indicates his or
her social position and at lifestyle as a behaviour typology of activity and time use patterns
(van Acker 2015). Specifically, the current study hypothesises that mode choice decisions
for short distances are not the result of the spur of the moment but rather a conscious long-
term decision to be a car driver, a cyclist, or a public transport user. Accordingly, the
current study adopts the perspective of lifestyle as a behavioural typology of mode choice
for short distances and elicits the lifestyle segmentation simultaneously with the lifestyle
expression.
When looking at lifestyle from a measurement perspective, the current study aims at
overcoming issues that characterise the vast majority of the literature analysing lifestyle
and travel behaviour. van Acker (2015) proposes an insightful review of the measurement
options and distinguishes the following approaches: (i) a socio-economic and demographic
approach that measures life stages and household composition; (ii) a psychographic life-
style approach that analyses personality traits and related motives and values of the
individual; (iii) a cultural lifestyle approach that shifts the focus on the community rather
than the individual; (iv) a sociographic lifestyle approach that moves the focus back
towards the individual and specifically his or her opinions and attitudes; (v) a mechanistic
lifestyle approach that considers lifestyle as a way of living; (vi) a post-structural lifestyle
approach that disconnects lifestyle and social structure; (vii) a geographic lifestyle
approach that combines information on the individual with spatial information on any
location that is important to the individual. Accordingly, the current study measures
lifestyle with a socio-economic and demographic approach while considering also the
mechanistic approach since, for example, deciding to be a cyclist is hypothesised to be a
way of life that relates to the socio-demographic characteristics and the activity patterns of
the individual and likely influences the choice of travel modes.
More relevantly from the measurement perspective, studies about lifestyle in the
transportation literature often rely on attitudinal data that are collected from surveys (e.g.,
Anable 2005; Schwanen and Mokhtarian 2005; Krizek 2006; Kitamura 2009; Ory and
Mokhtarian 2009; Sun et al. 2012; van Acker 2015) and generally follow a two-stage
approach where separation exists between the extraction of lifestyles and the correlation of
the lifestyle factors with travel behaviour (see, e.g., Walker and Li 2007). The two latter
propositions are significant limitations of existing studies on lifestyle: (i) attitudinal sur-
veys are often designed without a proper validation of the scales in the questionnaires and
are generally administered to convenience samples without statistical representativity of
the population, and obviously these two issues hinder the generalization of any result and
limit the validity of any conclusion on the lifestyles; (ii) assuming that lifestyle preferences
are error-free in a two-stage approach is clearly incorrect because the consequent mea-
surement error is a form of endogeneity that biases parameter estimates and thwarts the
Transportation
123
validity of any conclusion on the lifestyle manifestations. The current study overcomes
these two limitations by analysing mode choice decisions of a representative sample of the
population and estimating simultaneously the mode choice behaviour as an indicator for
the lifestyle construct and using this information for the definition of the lifestyle seg-
mentation. A latent class choice model is the most appropriate for this simultaneous
estimation since the hypothesis is that lifestyle preferences exist, these lifestyles are not
observable from the data, and individuals with different lifestyles have different mode
choice behaviour when travelling short distances (see, e.g., Walker and Li 2007).
Methods
Data about short trip chains of a representative sample of the Danish population were
obtained from the Danish National Travel Survey (TU, in Danish Transportvane-Un-
dersøgelsen) and a latent class choice model were used to verify the hypothesis that
lifestyle decisions influence mode choices for short distances in the Copenhagen Region.
Data
The TU survey collects information about the travel behaviour of a representative sample
of the Danish population between 10 and 84 years old via the administration of about 1000
interviews per month (about 80 % by telephone and about 20 % on the Internet) since
2006. The TU survey is administered by the Department of Transport of the Technical
University of Denmark with the support of an external consultant for the calibration of the
representative sample. The TU survey participants are extracted via a stratified random
procedure from the Danish Civil Registration System (in Danish, Det Centrale Person-
register) managed by the Danish National Board of Health (in Danish, Sundhedsstyrelsen)
with the objective of reaching representativity of the population as listed in the Danish
National Register managed by the Danish Census Bureau (in Danish, Danmark Statistik).
The Danish Data Protection Agency (in Danish, Datatilsynet) permits the use of sensitive
data for research purposes, namely names, addresses, and coordinates of the movements.
Approximately 95 % of the locations (e.g., home addresses, workplace addresses, trip
points) are coded geographically by the respondent with a ‘‘search and select’’ available in
the survey. Addresses are identified at the coordinate level in 98 % of the cases, and at the
zone level in 99.9 % of the cases, which implies that absolute confidentiality is guaranteed
prior to processing the data.
The 10,982 short trip chains analysed in the current study were the result of the
application of the following criteria: (i) the trip chains were below a distance threshold of
22 km, which constitutes the 95 % percentile of the trip chains by active travel mode in the
Copenhagen Region and accordingly is a realistic distance threshold to be considered for
curbing automobile use in short trips; (ii) the trip chains were performed by the population
over 18 years of age that constitutes the driving licensing age in Denmark (trip availability
considered the car availability in the household). The short trip chains contained detailed
information about: (i) the socio-economic-demographic characteristic of the 7958 indi-
viduals between 18 and 84 years old that travelled for short distances; (ii) the level-of-
service variables of the trip chains by walking, cycling, driving, being a passenger in a car,
and being a passenger in a public transport vehicle (e.g., bus, metro, train); (iii) the context
of the trip chain in terms of trip purpose, detailed location characteristics (e.g., station type,
Transportation
123
Table 1 Sample characteristics
Variable Categories (weighted)
Individual characteristics
Gender Male 48.3 % Female 51.7 %
Age 18–24 8.8 % 45–54 17.9 %
25–34 19.8 % 55–64 15.8 %
35–44 22.2 % 65 or older 15.5 %
Single Yes 28.9 % No 71.1 %
Chidren 0–4 Yes 13.9 % No 86.1 %
Children 5–9 Yes 14.6 % No 85.4 %
Children 10–15 Yes 15.8 % No 84.2 %
Occupation Student 10.3 % Retired 20.1 %
Employed 61.9 % Unemployed 3.1 %
Self-employed 4.6 %
Income (yearly) Mean 272,950 kr. Standard deviation 230,610 kr.
Bicycle Yes 80.0 % No 20.0 %
Number of cars None 31.2 % Two 13.7 %
One 53.7 % Three or more 1.4 %
Driving license Yes 83.4 % No 16.6 %
Public transport monthly pass Yes 18.3 % No 81.7 %
Parking availability at destination Yes 90.5 % No 9.5 %
Free parking at destination Yes 54.3 % No 45.7 %
Trip characteristics
Trip purpose Commute 20.7 % Shopping 31.6 %
Business 2.3 % Escorting 9.0 %
Leisure 26.0 % Other 10.5 %
Location Copenhagen centre 36.6 % Minor town 11.8 %
Copenhagen area 40.9 % Rural area 10.7 %
Transportation
123
Table 1 continued
Variable Categories (weighted)
Travel time (walking) Mean 23.40 min Standard deviation 23.40
Travel time (cycling) Mean 22.72 min Standard deviation 19.26 min
Travel time (driving) Mean 10.01 min Standard deviation 7.98 min
Travel time (car passenger) Mean 10.01 min Standard deviation 7.98 min
Access time (public transport) Mean 0.26 min Standard deviation 0.87 min
Waiting time (public transport) Mean 14.74 min Standard deviation 14.64 min
In-vehicle time (public transport) Mean 17.36 min Standard deviation 16.72 min
Number of transfers (public transport) Mean 0.37 Standard deviation 0.75
Travel cost Mean 18.13 kr Standard deviation 26.90 kr
Temperature Mean 10.28 C Standard deviation 7.52 C
Rain Yes 22.2 % No 77.8 %
Transportation
123
parking availability), weather conditions (e.g., temperature, rain); (iv) the weight of each
trip chain that guarantees the representativity of the sample given that the participation to
the TU survey is voluntary (62.5 % complete responses) and self-selection of population
strata is observed. The level-of-service variables were calculated by knowing the network
conditions at the time of the trip chain and by assuming shortest path choices for walking
and cycling, shortest path choices conditional on the congestion conditions for driving and
being a passenger in a car, and detailed indication of the route choices by public transport
as collected in the dedicated section of the TU survey (Anderson et al. 2014).
Table 1summarises the characteristics of the sample, corrected by the weights allowing
to achieve population representativity. The short trip chains in the Copenhagen region were
18.0 % by walking, 28.4 % by cycling, 39.3 % by driving, 6.1 % by being a passenger in a
car, and 8.2 % by being a passenger on a public transport vehicle. Remarkably, almost half
of the short trip chains were still done by car in a city like Copenhagen that offers plenty of
sustainable transport alternatives. The sample shows almost equal share of men and
women, almost equal proportion of age categories, representative percentage of children in
the various ages (0–4 are preschool children, 5–9 are children not travel independent, and
10–15 are children with initial travel independence), and representative variation across
individuals in terms of occupation and income. The sample also shows the characteristics
of the trip chains, with a heterogeneous composition according to time and cost, and the
share in terms of purpose and location with the majority in the centre of the city or in the
immediate neighbouring municipalities.
Model
A latent class choice model is the most suitable methodological approach to analyse the
effect of lifestyle on mode decisions for short distances. As previously clarified, the model
allows to simultaneously uncover lifestyle preferences that are not directly observable from
the data and to elicit mode choice preferences that are heterogeneous across the lifestyle
groups. Details about latent class choice models are provided by Gopinath (1995), Walker
(2001), and Greene and Hensher (2003).
The latent class choice model is composed of two parts: (i) a class membership model
that represents the probability of individual nto have lifestyle s, and (ii) a class specific
choice model that represents the probability of individual nwith a specific lifestyle sto
choose travel mode ifor short trip chain t. Given the characteristics X
n
of the individual
and the attributes X
i
of the travel modes, the probability of individual nto choose mode i
for short trip chain tis expressed as:
Pi
tXn;Xnit
jðÞ¼
X
S
s¼1
Pi
tXnit;sjðÞPsX
n
jðÞ ð1Þ
where P(s|X
n
) is the probability of individual nwith characteristics X
n
to have lifestyle s,
and P(i
t
|X
nit
,s) is the probability of individual n, conditional on having lifestyle s,to
choose mode iwith attributes X
nit
as perceived by individual nfor short trip chain t.It
should be noted that the probability of choosing mode ifor short trip chain tis equal to the
sum over all the Slifestyles of the products of the probability of the class specific choice
model (conditional on lifestyle s) and the probability of having that lifestyle.
In the current study, the class specific choice model is specified as an error component
logit that captures the correlation between alternative modes (i.e., active travel vs.
motorised travel) and the panel effect for multiple trip chains tbeing performed by
Transportation
123
individual nwith lifestyle s. Given five alternative modes available to individual n
(W=walk, C=bicycle, D=car driver, P=car passenger, B=public transport) per-
forming Tshort trip chains, the utility functions U
nits
of the travel modes ifor short trip
chain tof individual nhaving lifestyle sare expressed as follows:
UnWts ¼bsXnWt þrAgA
nþenWts
UnCts ¼bsXnCt þrAgA
nþenCts
UnDts ¼bsXnDt þrMgM
nþenDts
UnPts ¼bsXnPt þrMgM
nþenPts
UnBts ¼bsXnBt þrMgM
nþenBts
ð2Þ
where the error components g
n
A
and g
n
M
capture the correlation across active travel modes
Aand motorised travel modes Mas well as the panel effect across individuals n. The error
components g
n
A
and g
n
M
are i.i.d. normally distributed with mean equal to zero and variance
equal to one, the error terms e
nWts
,e
nCts
,e
nDts
,e
nPts
, and e
nBts
are i.i.d. extreme value
distributed across individuals, trip chains, and lifestyles, and the vectors g(=g
n
M
,g
n
A
) and e
(=e
nWts
,e
nCts
,e
nDts
,e
nPts
,e
nBts
) are independent (see Walker 2001). The column vectors
X
nWt
,X
nCt
,X
nDt
,X
nPt
and X
nBt
contain the attributes of the travel modes as perceived by
individual nfor trip chain t, and they are obviously independent of the lifestyle s. The
parameters to be estimated are the row vectors b
s
, which are specific to each lifestyle s, and
the scalars r
A
and r
M
that are equal across lifestyles sto impose a parsimonious specifi-
cation of the error structure and facilitate model identification (see Walker and Li 2007).
In the current study, the class membership model is specified as a logit model where the
utility function U
ns
of individual nhaving lifestyle sis:
Uns ¼dsþcsXnþens ð3Þ
where the vector X
n
contains the socio-economic-demographic characteristics of the
individuals n,d
s
is a class specific constant to be estimated, c
s
is a vector of class specific
parameters to be estimated, and e
ns
is an i.i.d. extreme value distributed error term. It
should be noted that the probabilistic nature of the class membership model allows for each
individual to have a different probability of having a different lifestyle s, and hence to have
multiple lifestyles in which one might be dominant because of a very high probability.
Given the specification of the two components of the latent class choice model, the
probability of individual nchoosing mode ifor short trip chain tconditional on having
lifestyle sis expressed as:
Pi
tXnit;s;bs;rA
;rM
¼ZY
T
t¼1
Pi
tXnit;s;g;bs;rA
;rM
fgðÞdg:ð4Þ
This is the product of the logit probability of each individual nchoosing mode ifor each
of Ttrip chains (where the number of trip chains per individual varies, thus the panel is
unbalanced), conditional on the unknown gand hence integrated over the distribution of g.
The probability of individual nhaving lifestyle sis expressed as:
PsX
n;ds;cs
j
ðÞ¼
exp dsþcsXn
ðÞ
PS
r¼1exp drþcrXn
ðÞ
:ð5Þ
Accordingly, the probability of individual nchoosing mode ifor short trip chain tis:
Transportation
123
Pi
tXn;Xnit;bs;rA
;rM
;ds;cs
¼X
S
s¼1
Pi
tXnit;s;bs;rA
;rM
PsX
n;ds;cs
j
ðÞð6Þ
and the log-likelihood is expressed as:
LL ¼X
tX
i
dnitxtln Pi
tXn;Xnit;bs;rA
;rM
;ds;cs
ð7Þ
where d
nit
is equal to one if individual nchooses mode ifor trip chain t(and 0 otherwise),
and x
t
is the weight of short trip chain t. The model is estimated via maximum likelihood
estimation, and numerical integration is used to evaluate the two-dimension integral in
Eq. (4). The model estimation produces simultaneously the parameter estimates for the
elements of the vectors b
s
,d
s
, and c
s
, and the scalars r
A
and r
M
, which allow evaluating the
different trade-offs made by individuals having different lifestyles. It should be noted that
the model is probabilistic in nature, namely each individual nhas a non-null probability to
have latent lifestyle s, and the estimate of the size of each lifestyle group is provided.
Moreover, the main issue with the model estimation is that the number of lifestyles cannot
be estimated endogenously, but the exogenous definition of the number Sof classes for the
estimation of different models can be performed and then the performances of the different
models can be compared. In the current study, latent class choice models were estimated
with Svarying from 2 to 6 and the number of classes was selected via a combination of
statistical information and interpretation of the estimation results.
Estimation results
Selection of the number of classes
As the number of classes cannot be estimated endogenously, latent class choice models were
estimated with the number of lifestyle varying between 2 and 6. It should be noted that also an
error component logit specification without segmentation of the individuals according to
lifestyle was estimated, and that the class specific and class membership models shared the
same specification in order to isolate the effect of the varying number of classes.
The selection of the number of classes relies on the easiness and logic of the behavioural
interpretation as well as the indices of model performances. The model performances of
the different models are presented in Table 2and the statistics supporting the selection of
Table 2 Performances of the estimated choice models
EC logit Latent class choice models
Number of classes 1 23456
Number of parameters 30 92 137 182 227 272
Log-likelihood at zero -17,675 -17,675 -17,675 -17,675 -17,675 -17,675
Log-likelihood at estimates -11,530 -10,743 -10,269 -10,030 -9875 -9804
Rho-bar squared 0.346 0.387 0.411 0.422 0.428 0.430
AIC -23,120 -21,670 -20,811 -20,425 -20,203 -20,153
BIC -23,339 -22,342 -21,812 -21,754 -21,861 -22,139
The values in bold are the best values for the Rho-bar squared, the AIC and the BIC measures of model
performance
Transportation
123
the number of classes were the rho-bar squared, the Akaike Information Criterion (AIC)
and the Bayesian Information Criterion (BIC). All the statistics are based on the same
principle of evaluating the goodness-of-fit of each model as measured by the log-likelihood
at estimates with respect to the parsimony as measured by the number of estimated
parameters. However, different statistics suggest that different models are preferable in
terms of goodness-of-fit versus parsimony. On the one hand, increasing the number of
parameters implies an increase in the goodness-of-fit when the evaluation is based on the
rho-bar squared and the AIC, although the rate of improvement in performances signifi-
cantly diminishes when estimating five and six latent classes. On the other hand, the same
phenomenon is not observed when the evaluation is based on the BIC, as this statistic
imposes a harsher penalty on the lack of parsimony. Given that the BIC suggests that the
behavioural interpretation appears easier and logical for class specific behaviour and class
membership of the 4-class choice model, and that the 4-class choice model gives the better
balance between goodness-of-fit and parsimony, estimates for this model are presented in
the remainder of this section.
For the sake of comparison, Table 3illustrates the estimates of the error component
logit model without latent lifestyle segmentation. Parameter estimates show that the
sample has comparable sensitivity to travel time by bicycle and car, significant sensitivity
to good weather conditions and hillyness when using active travel modes, and preferences
for specific modes according to the trip purpose.
The 4-class model: class specific behaviour
Table 4presents the parameter estimates of the class specific choice models with the same
specification of the error component logit without latent class segmentation. The two
parameters capturing the correlation across modes and the panel effect across individuals
are restricted to be equal across the four lifestyle groups for parsimony and identification
reasons, and are both significant to indicate that indeed unobservable similarities should
have been accounted for in the model specification.
It is evident that the 4-class choice model is better than the 1-class model not only from
the perspective of the goodness-of-fit, but also from the perspective of unravelling the
heterogeneity in the preferences across individuals. It should be noted that several
parameters are significant at the 0.05 and 0.10 confidence level (see the estimates in italic),
and also that several parameters are significantly different across classes according to a
Wald statistic test (see the note to the table). The examination of the estimated parameters
allows the definition of the lifestyle groups, especially when looking at the ratios between
the level-of-service estimates for the different travel modes across the different latent
lifestyles. Rates of substitution (Ben-Akiva and Lerman 1985) were calculated as the ratio
between two parameter estimates (e.g., in lifestyle group 1 the estimate for travel time by
bicycle is -0.129 and the estimate for travel time by car driver is -0.049, which implies
that the disutility for 1 min by bicycle is equal to the disutility for 2.63 min by car and
hence makes the car a lot better for this group).
Lifestyle group 1 is oriented towards the use of the automobile, as emerging from the
rate of substitution equal to 2.63 between car driver and bicycle parameters and equal to
2.30 between car passenger and bicycle parameters. Specifically, the individuals with this
lifestyle evaluate 1 min of travelling as car drivers as equal to 2.63 min of travelling as
cyclists, and hence they will likely never cycle to reach their destination. Also, this lifestyle
group exhibits high disutility for adverse weather conditions when walking and cycling and
for hillyness when needing a bicycle, expresses a positive preference for driving a car
Transportation
123
regardless of the purpose of the trip chain, and manifests a negative preference for cycling
especially for leisure and shopping purposes. Clearly, individuals with lifestyle 1 consider
the car as the fastest, cheapest, most comfortable and most convenient travel mode.
Lifestyle group 2 is oriented toward the use of the bicycle, as unravelling from the rate
of substitution equal to 2.33 between bicycle and car driver parameters and equal to 1.78
between bicycle and public transport parameters. Namely, the individuals with this life-
style perceive 1 min on the bicycle far better than the time spent in an automobile or a
public transport vehicle and rates of substitution almost inverse with respect to the indi-
viduals with lifestyle 1. Weather conditions are not significantly related to the choice of
walking and cycling, meaning that bicycle oriented individuals would not care whether it is
too hot, too cold or too wet when they need to travel. Also, hilliness is not significantly
Table 3 Estimates of the EC logit model (without latent lifestyle segmentation)
Variables Estimate T-statistic
Travel time—walk 20.085 29.73
Travel time—bicycle 20.075 29.25
Travel time—car driver 20.073 27.98
Travel time—car passenger 20.089 22.58
Waiting time—public transport 20.044 22.26
Access/egress time—public transport 20.054 22.54
In vehicle time—public transport 20.023 21.84
Number of transfers—public transport 20.736 210.65
Travel cost 20.052 21.92
Temperature—walk 0.028 2.25
Temperature—bicycle 0.077 5.67
Precipitation—walk 20.237 23.14
Precipitation—bicycle 20.141 22.35
Hilliness—bicycle 20.098 22.55
Parking availability—car driver 0.687 3.04
Monthly pass—public transport 0.897 3.89
Commuting purpose—bicycle 0.104 1.41
Commuting purpose—car driver 0.522 2.68
Commuting purpose—public transport 0.092 0.27
Leisure purpose—bicycle 20.815 22.93
Leisure purpose—car driver 20.405 21.83
Leisure purpose—public transport 20.452 21.40
Shopping purpose—bicycle 20.405 22.58
Shopping purpose—car driver 0.536 2.37
Alternative specific constant—walk 1.185 6.60
Alternative specific constant—car driver 21.482 28.22
Alternative specific constant—car passenger 22.597 213.97
Alternative specific constant—public transport 22.590 212.77
Standard deviation on active travel 1.090 11.25
Standard deviation on motorized travel 1.264 11.73
Estimates in bold and italic are significant at the 0.05 level, estimates in italic only are significant at the 0.10
level
Transportation
123
Table 4 Estimates of the class specific choice model
Variables Lifestyle independent Lifestyle 1 Lifestyle 2 Lifestyle 3 Lifestyle 4
Estimate T-statistic Estimate T-statistic Estimate T-statistic Estimate T-statistic Estimate T-statistic
Travel time—walk
a
20.129 22.99 20.074 21.54 20.058 22.03 20.085 21.71
Travel time—bicycle
a
20.129 23.36 20.054 23.81 20.087 22.10 20.072 22.01
Travel time—car driver
a
20.049 22.37 20.126 22.74 20.129 22.67 20.082 21.73
Travel time—car passenger
a
20.056 21.47 20.112 21.45 20.150 22.29 20.116 21.78
Waiting time—public transport
a
20.075 22.25 20.022 22.69 20.014 22.85 20.095 21.85
Access/egress time—public transport 20.059 21.83 20.052 24.72 20.033 23.28 20.091 22.66
In vehicle time—public transport 20.038 21.75 20.023 21.65 20.013 22.62 20.051 21.85
Number of transfers—public transport
a
21.157 24.23 20.416 21.69 20.188 21.89 22.165 24.47
Travel cost 20.077 21.51 20.072 22.55 20.032 21.44 20.073 21.65
Temperature—walk 0.029 1.01 0.017 1.77 0.030 0.90 0.040 1.69
Temperature—bicycle 0.205 2.26 0.017 0.38 0.126 1.88 0.088 5.11
Precipitation—walk 20.364 22.38 20.187 21.29 20.192 21.86 20.214 21.34
Precipitation—bicycle 20.343 22.02 20.151 20.89 20.162 20.83 20.254 21.58
Hilliness—bicycle
a
20.443 22.22 0.172 3.12 0.217 1.40 20.252 21.49
Parking availability—car driver 0.460 2.04 0.246 1.25 0.155 1.21 0.774 6.72
Monthly pass—public transport
a
0.359 1.56 0.252 1.59 1.295 10.87 0.414 3.48
Commuting purpose—bicycle
a
20.525 22.39 0.730 3.99 0.143 0.49 0.287 2.30
Commuting purpose—car driver
a
1.092 4.20 20.637 22.45 0.152 0.58 0.256 0.99
Commuting purpose—public transport
a
0.188 0.73 0.236 0.68 0.690 2.50 20.463 22.00
Leisure purpose—bicycle
a
21.671 22.67 20.465 21.34 20.791 21.70 20.838 22.27
Leisure purpose—car driver
a
0.623 2.35 20.502 21.03 20.698 22.10 21.016 22.75
Leisure purpose—public transport
a
20.553 20.98 20.211 20.35 0.907 2.56 20.728 21.81
Shopping purpose—bicycle
a
21.249 23.18 20.227 21.26 20.422 22.02 20.373 21.58
Shopping purpose—car driver
a
0.819 2.26 0.289 0.72 0.471 1.25 0.894 1.98
Transportation
123
Table 4 continued
Variables Lifestyle independent Lifestyle 1 Lifestyle 2 Lifestyle 3 Lifestyle 4
Estimate T-statistic Estimate T-statistic Estimate T-statistic Estimate T-statistic Estimate T-statistic
Alternative specific constant—walk
a
1.396 1.47 21.113 21.19 3.388 2.69 1.206 2.84
Alternative specific constant—car driver
a
2.168 2.17 25.479 27.15 24.500 23.65 0.677 1.35
Alternative specific constant—car passenger
a
1.039 1.18 25.434 27.18 24.224 23.43 20.895 21.80
Alternative specific constant—public transport
a
24.218 24.18 23.839 23.71 4.701 3.81 21.239 21.52
Standard deviation on active travel 1.112 10.82
Standard deviation on motorized travel 1.315 10.32
a
The parameters vary significantly across lifestyle groups (Wald statistic at the 0.10 confidence level), estimates in bold and italic are significant at the 0.05 level, estimates
in italic only are significant at the 0.10 level
Transportation
123
related to cycling, most likely because bicycle oriented individuals might enjoy the pos-
sibility of exercise that some hills offer. Commuting is the purpose that this lifestyle group
perceives as preferable when cycling, while non-significant relations are observed for
leisure and shopping trip chains. Evidently, individuals with lifestyle 2 consider the bicycle
as the fastest, most direct and most enjoyable travel mode.
Lifestyle group 3 is oriented toward the use of walk and public transport, as transpiring
from the rate of substitution equal to 2.14 between public transport and car driver
parameters, and equal to 1.43 between public transport and bicycle parameters. Clearly, the
individuals with this lifestyle perceive 1 min in a public transport vehicle better than the
time spent in a car, and slightly better than the time spent on a bicycle. Moreover, their
perception of the transfer penalty is lower than the one of the previous two lifestyle groups:
the penalty is 3.10 min per transfer for lifestyle 3, while it is equal to 4.31 min per transfer
for lifestyle 2 and 6.10 min per transfer for lifestyle 1. Individuals in the lifestyle group 3
have made the conscious decision of purchasing a monthly card to use public transport,
they are sensitive to adverse weather conditions when cycling, and even though they have
positive preferences for public transport commuting and leisure travel, they recognise the
convenience of the automobile for shopping trip chains. Noticeably, individuals with
lifestyle 3 consider walking and public transport as the preferable options in terms of time
saving and comfort.
Lifestyle group 4 does not appear to have a clear orientation towards a specific travel
mode, and the rates of substitution between bicycle and car driver parameters, bicycle and
car passenger parameters, and bicycle and walking parameters are in the proximity of the
unit. However, individuals in this lifestyle group have clearly a high negative sensitivity to
public transport as shown by the highest transfer penalty of 9.14 min per transfer and by
the highest rates of substitution of 3.27 between bicycle and public transport parameters
and 2.90 between car driver and public transport parameters. Individuals with this lifestyle
clearly dislike public transport, have comparable preferences for the other travel modes
regardless of being active or motorised and regardless of the trip purpose. Manifestly,
individuals with lifestyle 4 show an aversion for public transport and comparable values of
the other modes.
Summarising, lifestyle group 1 is automobile oriented, lifestyle group 2 is bicycle
oriented, lifestyle group 3 is walk and public transport oriented, and lifestyle group 4 is
public transport averse.
The 4-class model: class membership model
After presenting the estimates of the class specific choice models that allow illustrating the
heterogeneity of travel behaviour across individuals and labelling the latent lifestyle
groups, Table 5presents the estimates of the class membership choice model that allows
observing whether socio-economic-demographic characteristics of the individuals in the
sample are predictors of the latent lifestyle belonging. It should be noted that several
parameters are significant at the 0.05 and 0.10 confidence level (see respectively the
estimates in bold and italic, and in italic), and also that most parameters are significantly
different across classes according to a Wald statistic test (see the note to the Table 5).
Lifestyle group 1 is more likely to be composed of individuals who are male, are over
their thirties, are living with other adults in their households, and also have small children.
As they are car oriented, the probability of belonging to this lifestyle group logically
increases with an increase of the income and of the number of cars in the household, the
residence in the municipalities in proximity of Copenhagen rather than in the centre of
Transportation
123
Table 5 Estimates of the class membership model
Variables Lifestyle 1 Lifestyle 2 Lifestyle 3 Lifestyle 4
Estimate T-statistic Estimate T-statistic Estimate T-statistic Estimate T-statistic
Constant
a
20.617 24.09 20.550 21.82 1.426 2.97 20.259 20.86
Male
a
0.529 5.45 0.220 1.32 20.994 22.69 0.246 1.39
Age 18–30 (piecewise) 0.004 0.74 0.016 4.37 20.024 22.43 0.003 0.82
Age 31–60 (piecewise) 0.012 2.40 20.012 21.71 0.004 0.36 20.003 20.75
Age 60 plus (piecewise) 0.017 2.83 20.008 21.04 0.001 0.09 20.010 22.00
Adults over 18 years old
a
0.450 2.18 20.207 21.82 20.101 20.76 20.142 20.69
Children under 5 years old
a
1.615 3.29 20.277 22.52 20.971 22.58 20.367 22.31
Children 5–10 years old
a
1.300 1.96 0.216 2.01 20.781 22.26 20.734 22.09
Children 10–15 years old
a
1.145 1.56 0.176 1.71 20.847 21.85 20.473 21.78
Number of cars
a
0.625 3.15 20.436 23.70 20.378 21.41 0.190 1.50
Income
a
0.162 4.82 20.073 22.63 20.168 21.68 0.080 1.55
Copenhagen centre
a
20.502 21.53 0.385 2.95 0.241 2.21 20.125 21.11
Copenhagen area
a
0.659 1.92 20.161 21.07 0.169 1.45 20.668 21.91
Student
a
20.691 22.06 0.508 1.22 0.408 1.09 20.227 20.63
Self-employed
a
0.745 2.06 21.126 22.78 20.396 21.13 0.777 2.39
Retired
a
0.191 0.51 0.754 1.70 0.133 0.31 21.080 23.04
Unemployed
a
20.947 22.38 0.148 0.35 1.171 2.82 20.373 21.03
a
The parameters vary significantly across lifestyle groups (Wald statistic at the 0.10 confidence level), estimates in bold and italic are significant at the 0.05 level, estimates
in italic only are significant at the 0.10 level
Transportation
123
Copenhagen, and the occupation being self-employed or salary-employed (note that the
parameter associated with students is negative with respect to the salary-employed refer-
ence category). In other words, individuals with this lifestyle are more affluent, have
growing families, have developing careers, and have a residence outside the urban core.
Lifestyle group 2 is more probable to be made of individuals who are male, are in their
twenties, and are not living with other adults in their households but might have children.
As they are bicycle oriented, the likelihood of having this lifestyle reasonably decreases
with higher income and higher number of cars, increases with the residence being in the
city centre rather than the outskirts of the metropolitan area or the rural parts of the
Copenhagen region, and increases with the occupation being a student or a retired worker.
In other words, individuals with this lifestyle are both younger students or salary-employed
workers who enjoy the vibrant city centre and older retired workers who have grown up
children.
Lifestyle group 3 is more likely to be constituted by individuals who are female, and
whose family composition, with or without adults and with or without children in the
household, is not significantly correlated with the membership in this class. As they are
walk and public transport oriented, the probability of being in this lifestyle group increases
for residents of the city centre, students, salary-employed with respect to self-employed,
and unemployed. In other words, it seems a bit more complex to profile the group that is
however mainly constituted by younger female students or salary-employed workers
without small children.
Lifestyle group 4 is more probable to be comprised of individuals of both genders who
may be in their twenties and thirties, and do not have small children. As they are public
transport averse, the likelihood of having this lifestyle does not relate to having higher
income or higher number of cars, decreases in the outskirts of the metropolitan area while
not significant difference exists for residence in either the city centre or the rural areas of
the Copenhagen region, and decreases for retired workers but increases for self-employed
workers. In other words, this lifestyle group is heterogeneous in income and residence
location and most likely has the highest flexibility in choosing between bicycle and
automobile.
The last piece of information estimated with the class membership model is the com-
position of the four lifestyle groups over the sample, calculated as the average probability
of belonging to each latent class. These probabilities are fairly split in the individuals in the
sample: 36 % for lifestyle 1, 27 % for lifestyle 2, 15 % for lifestyle 3, and 22 % for
lifestyle 4.
Summary and conclusions
The current study has verified the hypothesis that lifestyle influences mode choices for short
distances in the Copenhagen region and affects not only the choice but also the perception of
travel modes. The contribution of this study with respect to previous studies about short trips
lies in the consideration of the heterogeneity in individual preferences, especially when
comparing parameter estimates for travel time variables, and the relevance of lifestyle
decisions on short-term choices, especially on the basis of a representative sample of trav-
ellers. Very relevantly, the findings of this study highlight that analysing the potential for
switching from the automobile to sustainable travel modes while considering homogeneous
population is a simplistic assumption (see, e.g., Banister 2008; Monzon 2011).
Transportation
123
The findings of this study highlight that the population in the Copenhagen region is
composed of four heterogeneous types of individuals: (i) car oriented individuals are likely
more affluent and careerist individuals who have made the conscious decision of buying a
car, quite an expensive endeavour in Denmark given the high registration tax, and most
likely use the car in every trip chain regardless of the distance; (ii) bicycle oriented
individuals are likely younger and at a different stage in life with less children and more
interest in the vibrant city centre, and most likely they use the bicycle in most of the trip
chains regardless of the distance; (iii) walk and public transport oriented individuals have
made the conscious decision of not using the automobile for their trip chains and at the
same time do not exhibit a strong negative preference for the bicycle even though they
prefer public transport for reaching their farthest destinations; (iv) public transport averse
individuals form the most heterogeneous group and are the most flexible to the use of either
the bicycle or the automobile depending on the lifestyle stage.
When thinking about promoting sustainable travel modes and even active travel modes,
the findings of this study suggest that individuals in lifestyle groups 3 and 4 are the most
likely to be swayed to move towards these travel options. While lifestyle group 3 already
prefers walking as an alternative, it is evident that the probability of using the bicycle for
longer distances than the walkable ones would not be too low when looking at the rates of
substitution and the lower emphasis on weather conditions. Even more relevantly, while
lifestyle group 4 really dislikes public transport, it is clear that the probability of using the
bicycle is, at least, equal to the one of using the car when considering travel time (ceteris
paribus). Accordingly, evaluating the potential of shifting from the automobile to sus-
tainable and active travel modes should consider that any intervention should be directed
towards these lifestyle groups in primis. Policies against the car use, such as congestion
charging and high registration taxes, might shift individuals from having more disposable
income and more cars (as having automobiles becomes more expensive) and possibly
changing their lifestyle towards a sustainable one. These policies should also be accom-
panied by measures that make the travel time of modes alternative to the car much more
convenient, as car oriented individuals penalise heavily the time spent on anything other
than their automobiles.
When thinking about infrastructure improvements, for example looking at the heavy
investments of the Copenhagen municipality in improving bicycle infrastructure, the
reduction of the cycling travel time is assumed to make cycling more attractive with
respect to alternative travel modes. However, the estimated rates of substitution of the
travel time parameters for the car oriented lifestyle group show that the reduction should be
by a factor of at least two in order for the individuals within this group to even consider the
bicycle as a plausible alternative. The rates of substitution of the travel time parameters for
the bicycle oriented lifestyle group illustrate that the reduction would not have any impact
on the individuals in this group, since they already perceive the bicycle as the fastest and
most convenient travel mode. Instead, the rates of substitution of the travel time parameters
for the public transport oriented and averse lifestyle groups suggest that the balance could
be moved towards cycling even with modest reductions.
Limitations to the study are recognised. Firstly, the travel survey data are collected by
identifying single modes for the trip chains and multimodality does not appear in the data
also because of the specific nature of the region where individuals select a mode for the
entire tour rather than a combination of modes (especially for short distances). Secondly,
the specificity of the fully integrated public transport system in the Copenhagen Region
(with a zone fare system that allows for seamless transfer between metro, buses and trains)
Transportation
123
makes the study less general with respect to regions where the public transport modes
compete.
Further avenues for research are identifiable. Firstly, the current study estimates a latent
class choice model on the basis of traditional travel survey data and hence does not look at
the angle of attitudes and perceptions. An extension could involve the preparation of
questionnaires that would use validated scales and would capture the psychological aspects
behind the lifestyles observed from the travel survey information. Latent variable models
integrated with latent class models could be estimated according to the generalised
framework introduced by Walker (2001). Secondly, the current study assumes utility
maximization for all the lifestyle groups and hence does not open to different behavioural
paradigm. An extension could entail the estimation of a latent class model with different
formulation of the class specific choice models (e.g., regret minimization, lexicographic)
according to the framework introduced by Hess et al. (2012).
Acknowledgments The authors gratefully acknowledge the Danish Road Directorate for the financial
support of the project ‘‘Effects of cycling policies’’ that the current study is part of.
References
Anable, J.: ‘Complacent car addicts’ or ‘aspiring environmentalists’? Identifying travel behaviour segments
using attitude theory. Transp. Policy 12, 65–78 (2005)
Anderson, M.K., Nielsen, O.A., Prato, C.G.: Multimodal route choice models of public transport passengers
in the Greater Copenhagen Area. EURO J. Transp. Logist. (2014). doi:10.1007/s13676-014-0063-3
Banister, D.: The sustainable mobility paradigm. Transp. Policy 15, 73–80 (2008)
Ben-Akiva, M.E., Lerman, S.: Discrete Choice Analysis. The MIT Press, Cambridge (1985)
Bina, M., Graziano, F., Bonino, S.: Risky driving and lifestyles in adolescence. Accid. Anal. Prev. 38,
472–481 (2006)
Borken-Kleefeld, J., Fuglestvedt, J.S., Berntsen, T.: Mode, load, and specific climate impact from passenger
trips. Environ. Sci. Technol. 47, 7608–7614 (2013)
Carse, A., Goodman, A., Mackett, R.L., Panter, J., Ogilvie, D.: The factors influencing car use in a cycle-
friendly city: the case of Cambridge. J. Transp. Geogr. 28, 67–74 (2013)
Choo, S., Mokhtarian, P.L.: What type of vehicle do people drive? The role of attitude and lifestyle in
influencing vehicle type choice. Transp. Res. A 38, 201–222 (2004)
de Hartog, J.J., Boogaard, H., Nijland, H., Hoek, G.: Do the health benefits of cycling outweigh the risks?
Environ. Health Perspect. 118, 1109–1116 (2010)
de Nazelle, A., Morton, B.J., Jerrett, M., Crawford-Brown, D.: Short trips: an opportunity for reducing
mobile-source emissions? Transp. Res. D 15, 451–457 (2010)
de Nazelle, A., Nieuwenhuijsen, M.J., Anto
´, J.M., Brauer, M., Briggs, D., Braun-Fahrlander, C., Cavill, N.,
Cooper, A.R., Desqueyroux, H., Fruin, S., Hoek, G., Panis, L.I., Janssen, N., Jerrett, M., Joffe, M.,
Andersen, Z.J., van Kempen, E., Kingham, S., Kubesch, N., Leyden, K.M., Marshall, J.D., Matamala,
J., Mellios, G., Mendez, M., Nassif, H., Ogilvie, D., Peiro
´, R., Pe
´rez, K., Rabl, A., Ragettli, M.,
Rodrı
´guez, D., Rojas-Rueda, D., Ruiz, P., Sallis, J.F., Terwoert, J., Toussaint, J.F., Tuomisto, J.,
Zuurbier, M., Lebret, E.: Improving health through policies that promote active travel: a review of
evidence to support integrated health impact assessment. Environ. Int. 37, 766–777 (2011)
Fan, Y., Khattak, A.: Time use patterns, lifestyles, and sustainability of nonwork travel behavior. Int.
J. Sustain. Transp. 20, 26–47 (2012)
Forsey, D., Nurul Habib, K., Miller, E.J., Shalaby, A.: Temporal transferability of work trip mode choice
models in an expanding suburban area: the case of York Region, Ontario. Transp. A 10, 469–482
(2014)
Frank, L., Stone, B.J., Bachman, W.: Linking land use with household vehicle emissions in the central Puget
sound: methodological framework and findings. Transp. Res. D 5, 173–196 (2000)
Frenkel, A., Bendit, E., Kaplan, S.: The linkage between the lifestyle of knowledge-workers and their intra-
metropolitan residential choice: a clustering approach based on self-organizing maps. Comput. Env-
iron. Urban Syst. 39, 151–161 (2013a)
Transportation
123
Frenkel, A., Bendit, E., Kaplan, S.: Residential location choice of knowledge-workers in a ‘‘startup
metropolis’’: the role of amenities, workplace and lifestyle. Cities 35, 33–41 (2013b)
Fuglestvedt, J.S., Shine, K.P., Berntsen, T., Cook, J., Lee, D.S., Stenke, A., Skeie, R.B., Velders, G.J.M.,
Waitz, I.A.: Transport impacts on atmosphere and climate: metrics. Atmos. Environ. 44, 4648–4677
(2010)
Gopinath, D.A.: Modeling Heterogeneity in Discrete Choice Processes: Application to Travel Demand.
Ph.D. Dissertation, Massachusetts Institute of Technology, Cambridge, MA (1995)
Grabow, M., Spak, S., Holloway, T., Stone, B., Mednick, A., Patz, J.: Air quality and exercise-related health
benefits from reduced car travel in the Midwestern United States. Environ. Health Perspect. 120, 68–76
(2012)
Greene, W.H., Hensher, D.A.: A latent class model for discrete choice analysis: contrasts with mixed logit.
Transp. Res. B 37, 681–698 (2003)
Guerra, E.: The geography of car ownership in Mexico City: a joint model of households’ residential
location and car ownership decisions. J. Transp. Geogr. 43, 171–180 (2015)
Hertel, O., Hvidberg, M., Ketzel, M., Storm, L., Stausgaard, L.: A proper choice of route significantly
reduces air pollution exposure—a study on bicycle and bus trips in urban streets. Sci. Total Environ.
389, 58–70 (2008)
Hess, S., Stathopoulos, A.: A mixed random utility—random regret model linking the choice of decision
rule to latent character traits. J. Choice Model 9, 27–38 (2013)
Hess, S., Stathopoulos, A., Daly, A.J.: Allowing for heterogeneous decision rules in discrete choice models:
an approach and four case studies. Transportation 39, 565–591 (2012)
Kim, S., Ulfarsson, G.F.: Curbing automobile use for sustainable transportation: analysis of mode choice on
short home-based trips. Transportation 35, 723–737 (2008)
Kitamura, R.: Life-style and travel demand. Transportation 36, 679–710 (2009)
Krizek, K.J.: Lifestyles, residential location decisions, and pedestrian and transit activity. Transp. Res. Rec.
1981, 171–178 (2006)
Krizek, K.J., Waddell, P.: Analysis of lifestyle choices: neighborhood type, travel patterns, and activity
participation. Transp. Res. Rec. 1807, 119–128 (2002)
Lindsay, G., Macmillan, A., Woodward, A.: Moving urban trips from cars to bicycles: impact on health and
emissions. Aust. N. Z. J. Public Health 35, 54–60 (2011)
Loukopoulos, P., Ga
¨rling, T.: Are car users too lazy to walk? The relationship of distance thresholds for
driving to the perceived effort of walking. Transp. Res. Rec. 1926, 206–211 (2005)
Mabit, S.L., Fosgerau, M.: Demand for alternative-fuel vehicles when registration taxes are high. Transp.
Res. D 16, 225–231 (2011)
Mackett, R.: Why do people use their cars for short trips? Transportation 30, 329–349 (2003)
Maibach, E., Steg, L., Anable, J.: Promoting physical activity and reducing climate change: opportunities to
replace short car trips with active transportation. Prev. Med. 49, 326–327 (2009)
Monzon, A., Vega, L.A., Lopez-Lambas, M.E.: Potential to attract drivers out of their cars in dense urban
areas. Eur. Transp. Res. Rev. 3, 129–137 (2011)
Noland, R., Park, H., Von Hagen, L.A., Chatman, D.G.: A mode choice analysis of school trips in New
Jersey. J. Transp. Land Use 7, 111–133 (2014)
Ory, D., Mokhtarian, P.L.: Modeling the structural relationships among short-distance travel amounts,
perceptions, affections, and desires. Transp. Res. A 43, 26–43 (2009)
Paulssen, M., Temme, D., Vij, A., Walker, J.L.: Values, attitudes and travel behavior: a hierarchical latent
variable mixed logit model of travel mode choice. Transportation 41, 873–888 (2014)
Piatkowski, D.P., Krizek, K.J., Handy, S.L.: Accounting for the short term substitution effects of walking
and cycling in sustainable transportation. Travel Behav. Soc. 2, 32–41 (2015)
Pinjari, A.R., Pendyala, R.M., Bhat, C.R., Waddell, P.A.: Modeling the choice continuum: an integrated
model of residential location, auto ownership, bicycle ownership, and commute tour mode choice
decisions. Transportation 38, 933–958 (2011)
Pucher, J., Komanoff, C., Schimek, P.: Bicycling renaissance in North America? Recent trends and alter-
native policies to promote bicycling. Transp. Res. A 33, 625–654 (1999)
Rich, J., Prato, C.G., Hels, T., Lyckegaard, A., Kristensen, N.B.: Analyzing the relationship between car
generation and severity of motor-vehicle crashes in Denmark. Accid. Anal. Prev. 54, 81–89 (2013)
Rietveld, P.: Non-motorised modes in transport systems: a multimodal chain perspective for the Nether-
lands. Transp. Res. D 5, 31–36 (2000)
Rojas-Rueda, D., de Nazelle, A., Tainio, M., Nieuwenhuijsen, M.J.: The health risks and benefits of cycling
in urban environments compared with car use: health impact assessment study. Br. Med. J. 343, d4521
(2011)
Transportation
123
Salomon, I.: Life Style as a Factor in Explaining Travel Behavior. Ph.D. Dissertation, Massachusetts
Institute of Technology, Cambridge, MA (1980)
Salomon, I., Ben-Akiva, M.E.: The use of the life-style concept in travel demand models. Environ. Plan. A
15, 623–638 (1983)
Scheepers, E., Wendel-Vos, W., van Kempen, E., Panis, L.I., Maas, J., Stipdonk, H., Moerman, M., den
Hertog, F., Staatsen, B., van Wesemael, P., Schuit, J.: Personal and environmental characteristics
associated with choice of active transport modes versus car use for different trip purposes of trips up to
7.5 kilometers in the Netherlands. PLoS One 8, e73105 (2013)
Scheiner, J., Holz-Rau, C.: Travel mode choice: affected by objective or subjective determinants? Trans-
portation 34, 487–511 (2007)
Schepers, J.P., Heinen, E.: How does a modal shift from short car trips to cycling affect road safety? Accid.
Anal. Prev. 50, 1118–1127 (2013)
Schwanen, T., Mokhtarian, P.L.: What affects commute mode choice: neighborhood physical structure or
preferences toward neighborhoods? J. Transp. Geogr. 13, 83–99 (2005)
Sigurdardottir, S.B., Kaplan, S., Møller, M., Teasdale, T.: Understanding adolescents’ intentions to com-
mute by car or bicycle as adults. Transp. Res. D 24, 1–9 (2013)
Smith, B., Olaru, D.: Lifecycle stages and residential location choice in the presence of latent preference
heterogeneity. Environ. Plan. A 45, 2495–2514 (2013)
Sun, Z., Arentze, T., Timmermans, H.: A heterogeneous latent classmodel of activity rescheduling, route choice
and information acquisition decisions under multiple uncertain events. Transp. Res. C 25, 46–60 (2012)
van Acker, V.: Lifestyle and modal choices: Defining, measuring and using the ‘lifestyle’ concept. Pro-
ceedings of the 94th Annual Meeting of the Transportation Research Board, Washington, D.C (2015)
van Acker, V., Mokhtarian, P.L., Witlox, F.: Car availability explained by the structural relationships
between lifestyles, residential location, and underlying residential and travel attitudes. Transp. Policy
35, 88–99 (2014)
Vij, A., Carrel, A., Walker, J.L.: Incorporating the influence of latent modal preferences on travel mode
choice behaviour. Transp. Res. A 54, 164–178 (2013)
Vovsha, P., Vyas, G., Givon, D., Birotker, Y.: Individual mobility attributes and their impact on modality
style comparison across three population sectors in Jerusalem, Israel. Transp. Res. Rec. 2382, 132–141
(2013)
Walker, J.L.: Extended Discrete Choice Models: Integrated Framework, Flexible Error Structures, and
Latent Variables. Ph.D. Dissertation, Massachusetts Institute of Technology, Cambridge, MA (2001)
Walker, J.L., Li, J.: Latent lifestyle preferences and household location decisions. J. Geogr. Syst. 9, 77–101
(2007)
Zhou, J.: From better understandings to proactive actions: housing location and commuting mode choices
among university students. Transp. Policy 33, 166–175 (2014)
Carlo Giacomo Prato is currently Professor in Transport Engineering at the School of Civil Engineering of
the University of Queensland. His expertise in advanced modelling and data mining enable him to solve
challenges in representing travel behaviour in complex transport systems. Among his recent research
endeavours are the analysis of choice behaviour of cyclists, drivers and public transport users as well as the
evaluation of the value of time and the rates of substitutions for different parts of their journeys. His
contributions have been published in prestigious peer-reviewed journals and international conferences, and
his activities in the research community include being on the Editorial Board of prestigious journals like
Injury Prevention and Transportation Research Part F: Traffic Psychology and Behaviour.
Katrı
´n Halldo
´rsdo
´ttir has received her Ph.D. from the Doctorate School of the Department of Transport at
the Technical University of Denmark. Her dissertation focused on cyclists’ behaviour from both a mode
choice and a route choice perspective, and her contributions vary from data collection and processing
techniques to mode choice models for short trips, and from choice set generation for bicycle trips to route
choice models for commuting cyclists.
Otto Anker Nielsen is Professor in Transport Modelling at the Technical University of Denmark. His vast
experience with transport modelling generates from a vastly demonstrated talent in producing theoretical
contributions and a parallel ability in showing how these innovative models can be applied to large-scale
problems. He leads extensive transport modelling projects like the Danish National Transport Model and the
latest version of the Transtools model for all of Europe, and his contributions have been published in
prestigious peer-reviewed journals and international conferences.
Transportation
123
A preview of this full-text is provided by Springer Nature.
Content available from Transportation
This content is subject to copyright. Terms and conditions apply.
