Exploring Individual Preferences for Mobility as a Service Plans: A Mixed Methods Approach

Abstract

As the number of Mobility as a Service (MaaS) applications have started to grow, it is of high importance to gain a deeper understanding of the different aspects of this concept. With many different business models and product designs emerging, there is still little knowledge on what consumer preferences are, and how users navigate between various MaaS offerings. With this in mind, the main objective of this paper is to contribute to research about user preferences for MaaS products and more specifically for plans. In doing so, the paper aims to provide insights about what modes and features people prefer in their plans and what individual characteristics have a significant effect on choosing across MaaS plans. Taking London as a case study, an explanatory sequential mixed methods research design is used, in which the quantitative phase is carried out first, followed by the qualitative phase to help explain the quantitative results and provide additional insights. The quantitative phase includes an online survey and stated preference experiments followed by a MaaS plan choice model, while the qualitative comprises of in-depth interviews and thematic analysis. The results of the study provide interesting insights to consumer behavior and can be valuable to the private and public sector as well as the research community.

Full text

Matyas and Kamargianni
1
1
Working Paper
Exploring Individual Preferences for Mobility
as a Service Plans: A Mixed Methods Approach
UCL, Central House, 14 Upper Woburn Place, WC1H 0NN,
London, UK
Melinda Matyas, Corresponding Author
MaaSLab, Energy Institute, University College London
Email: Melinda.matyas.13@ucl.ac.uk
Maria Kamargianni
MaaSLab, Energy Institute, University College London
Email: m.kamargianni@ucl.ac.uk
MaaSLab-WP-18-01
July, 2018
To cite this paper: Matyas, M. and Kamargianni, M. (2018) Exploring Individual Preferences for
Mobility as a Service Plans: A Mixed Methods Approach. MaaSLab Working Paper Series Paper No.
18-02.
MaaSLab-WP-18-01

Matyas and Kamargianni
2
2
Abstract:
As the number of Mobility as a Service (MaaS) applications have started to grow, it is of high
importance to gain a deeper understanding of the different aspects of this concept. With many different
business models and product designs emerging, there is still little knowledge on what consumer
preferences are, and how users navigate between various MaaS offerings. With this in mind, the main
objective of this paper is to contribute to research about user preferences for MaaS products and more
specifically for plans. In doing so, the paper aims to provide insights about what modes and features
people prefer in their plans and what individual characteristics have a significant effect on choosing
across MaaS plans. Taking London as a case study, an explanatory sequential mixed methods research
design is used, in which the quantitative phase is carried out first, followed by the qualitative phase to
help explain the quantitative results and provide additional insights. The quantitative phase includes
an online survey and stated preference experiments followed by a MaaS plan choice model, while the
qualitative comprises of in-depth interviews and thematic analysis. The results of the study provide
interesting insights to consumer behavior and can be valuable to the private and public sector as well
as the research community.
Keywords: Mobility as a Service, MaaS, MaaS products, plans, mixed methods, bundles, choice
model, interview

Matyas and Kamargianni
3
3
1 Introduction
Riding the wave of technological advancements, the widespread availability of ICT devices and the
sharing economy, the past decade has seen the introduction of a number of new mobility services. Car
sharing, bike sharing, on demand transport, and ridehailing are now commonplace in many urban
areas. Additionally, the next generation of technological innovations such as autonomous vehicles and
drones are now being tested. With the number of overall transport options available to travelers
continuously growing, it is increasingly difficult for users to navigate through the silos of different
information sources, mobile applications, tickets and journey planners that are necessary for them to
get around (Kamargianni et al., 2015). The need for a single, integrated, user-friendly system has led
to the birth of the Mobility as a Service (MaaS) concept, which aims to decrease the pain points that
result from multimodal journeys.
The definition of MaaS has been widely debated, but in this paper, we follow the definition provided
by MaaSLab (2018), according to which “Mobility-as-a-Service is a user-centric, intelligent mobility
management and distribution system, in which an integrator brings together offerings of multiple
mobility service providers, and provides end-users access to them through a digital interface, allowing
them to seamlessly plan and pay for mobility.” From the users’ point of view, the service is accessible
through a digital interface or smartphone app, where they can purchase MaaS products, such as
monthly subscription plans (Kamargianni, et al., 2016; Hietanen, 2016; Hensher, 2017). MaaS
subscription plans build on the concept of product bundling, which is the practice of marketing and
selling two or more products or services in a package for a special price (Guiltinan, 1987). Packaging
transport services via MaaS may be new, but the method is analogous to bundles in other sectors, such
as telecommunication (e.g. TV + Internet + Landline plans).
The MaaS concept is still not at its maturity and there are only a handful of applications. Commercial
applications are available in Helsinki and the West Midlands and Antwerp run by Whim from MaaS
Global, while pilots have been tested in Sweden, Austria and Germany (for detailed reviews of MaaS
and MaaS-like services, please refer to the works from Kamargianni, et al., 2016; Jittrapirom, et al.,
2017; Magoutas, et al., 2017; Georgakis, et al., 2018). The available services are frequently amending
their product offerings, which suggests that there is still much to learn about the exact demand and
user preferences for these. There is a small, but growing number of research papers that tackle this
topic (Matyas & Kamargianni, 2017; Ho, et al., 2017; González Alonso, et al., 2017; Sochor, et al.,
2015). However, the user preferences for different product attributes and the behavioral and societal
barriers to consumer adoption are still largely unknown.
With this in mind, the objective of this paper is to contribute to the understanding of user preferences
for MaaS plans. More specifically, the two main research questions are as follows: (1) What modes
and features do people favor in their MaaS plans and how do they evaluate between them? (2) What
individual characteristics have a significant effect on choosing MaaS plans?
To answer these questions an explanatory sequential mixed methods research design is used. This two-
step approach involves collecting and analyzing quantitative data first, followed by an in-depth
qualitative element. In the first, quantitative, phase of the study, an online survey, including a stated
preference experiment about MaaS plans is carried out. The data is analyzed using a mixed logit model
(MMNL) to explore individuals’ preferences towards the various elements of MaaS plans. The second,
qualitative, phase is based on in-depth semi-structured interviews with survey participants and the data
is analyzed through thematic analysis. The qualitative analysis helps explain the mechanisms or
reasons behind the trends and relationships observed in the quantitative data and also provides
additional insights. The case study city is London.
The remainder of the paper is structured as follows. Section 2 provides a literature review about
existing research on MaaS and mixed methods research within transport demand analysis in order to

Matyas and Kamargianni
4
4
demonstrate the contribution of this paper. Section 3 gives an overview of the methodology and
describes the design quantitative data collection tool. Section 4 provides the quantitative data analysis
and model results, while section 5 describes the qualitative data collection, analysis and results.
Finally, section 6 provides some discussion and conclusions.
2 Literature
2.1 Mobility as a Service Literature
Due to the relative novelty of the Maas concept, the available papers on the topic are fairly limited.
However, as the concept gains wider acceptance, the number of both academic and grey literature
getting published each year is constantly growing. The topics explored include business models
(Kamargianni & Matyas, 2017; Ebrahimi, et al., 2018), impacts on specific transport modes (Hensher,
2017; Smith, et al., 2017), end user demand (Sochor, et al., 2015; Ho, et al., 2017; Matyas &
Kamargianni, 2017), governmental role (Heikkilä, 2014), topology of services (Kamargianni, et al.,
2016; Jittrapirom, et al., 2017; Sochor, et al., 2017), issues around implementation of MaaS (Mulley,
2017; Wong, et al., 2017). Most existing literature consists of thought pieces and reviews. Some
examples from academic literature include studies, which review MaaS-like, integrated services in
order to highlight key characteristics (Jittrapirom, et al., 2017). Others, take it one step further and
attempt to classify these services based on the level of integration (Kamargianni, et al., 2016; Sochor,
et al., 2017). There are also studies, which provide thoughts on one aspect of MaaS, such as the
conditions of its implementation (Li & Voege, 2017) or the way various modes could be offered under
MaaS (Hensher, 2017). Grey literature also contributes a number of thought pieces and reviews
(Datson, 2016; Cubic Transportation Systems, 2016).
Even though most studies available are not based on empirical findings, there are a handful of such
papers available. These take either quantitative or qualitative approaches, or a combination of the two.
Looking at those that collect and analyze quantitative data, these studies use stated preference methods
to analyze demand for- and the effect of MaaS (Matyas & Kamargianni, 2017; Ho, et al., 2017;
González Alonso, et al., 2017). Examining these studies in more detail, González Alonso, et al. (2017)
focus on whether MaaS could create modal shift. They do this, by carrying out an on-line survey in
Amsterdam with 797 respondents, in which participants are asked to choose transport modes in nine
hypothetical situations. Using the mode choices, a modal portfolio is created for each respondent. Next,
for each modal portfolio, the ‘MaaS-prone’ behavior is investigated using basic statistical methods,
based on factors such as public transport subscription, mobility app usage and opinion towards
payment via applications. The study concludes, that multimodal users are most prone to adopt MaaS.
One interesting element of this study, is that it does not directly ask respondents anything about MaaS.
Rather, they use certain characteristics of individuals, namely public transport subscription, mobility
app usage and opinion towards payment via applications, as a proxy of their willingness to adopt MaaS.
In contrast, other studies collect data when respondents are directly asked about their preferences for
MaaS (Matyas & Kamargianni, 2017; Ho, et al., 2017). As the Matyas and Kamargianni (2017) study
is based on the same quantitative data collection as this paper, this will not be discussed further in this
literature review. Looking at another study, Ho et al. (2017) also design a stated preference experiment,
which includes asking respondents directly about their preferences for MaaS products. The SP
presented respondents with a choice between two fixed plans, a pay-as-you-go and a no choice (status
quo) option. The modes that were included in the plans are public transport, car sharing, taxi and
Uberpool. The survey was carried out in Sydney, Australia as a computer assisted personal interview
(CAPI) with a sample of 252 people. To analyze the collected data, discrete choice models were used
to get willingness to pay values for the different modes.
Turning to the studies using qualitative methods, these use individual meetings and semi-structured
interviews to examine various aspect of MaaS (Smith, et al., 2017; Smits, et al., 2017). Both studies

Matyas and Kamargianni
5
5
are more focused on the supply side, and conduct interviews with stakeholders in the MaaS ecosystem.
Smith et al. (2017a) conduct interviews with 19 actors in order to explore how MaaS could develop
and how the future of public transport could be affected; while Smith et al (2017b) carry out 10
interviews to identify aspect that are important when procuring MaaS. Finally, there are those studies
that use a mixed methods approach to explore various aspects of MaaS. A number of studies have
resulted from the UbiGo MaaS field trial (Sochor, et al., 2015; Karlsson, et al., 2016). In this trial,
public transport, car sharing, taxi, bike sharing and car rentals were offered to users as subscription
plans. 89 households with 195 users subscribed for monthly plans including a personalized
combination of- and credit for- the various travel services. The prepaid tailored monthly plans were
determined in time or distance for each mode and the combined subscription was cheaper than each
element individually. Credit could be topped up or rolled over and subscriptions modified. A mobility
broker handled everything for the users to make it a seamless experience. As this was an actual trial,
different provision structures/prices could not be tested (as it is possible the case of SPs). However,
the project did result in a number of studies on various elements of MaaS. Sochor et al. (2015) used
information collected through the pilot to identify matches and mismatches between the expectation
and experiences of three stakeholder groups: the users of the service, the commercial actors and
society. Karlsson et al. (2016) used a mixed methods data collection, which involved questionnaires,
travel diaries, interviews and focus groups with participants of the trial. The study involved basic
statistical analysis for the quantitative data and thematic analysis and selection of quotes for the
qualitative element. The qualitative and quantitative data were analyzed alongside each other. Their
findings indicate an overall positive outcome from the trial, however a number of barriers were also
recognized.
Overall, there is still only a limited amount of academic literature on MaaS, especially those that focus
on the end user and use empirical evidence. This study aims to contribute to these studies. It adds to
the existing literature by taking a mixed methods approach that uses both discrete choice models
alongside in-depth interviews. It also provides insights into a new city, one, that has very different
characteristics that the previously explored ones.
2.2 Mixed Methods in Travel Demand Analysis
Historically quantitative and qualitative methods were seen as incompatible approaches. Researchers
would frequently advocate one of the two methods and disparage the value of the alternative one
(Johnson & Onwuegbuzie, 2004; Tashakkori & Teddlie, 2010). However, since the 1980s
1
, studies
have been applying these methods together with increasing frequency (Creswell, 2003; Creswell &
Plano Clark, 2007).Using a combination of quantitative and qualitative elements, falls under the
mixed-method approach, in which multiple research methodologies are combined in one study. Mixed-
method approaches have been shown to provide significant added value to studies, especially when
the phenomenon under study is complex (Morse, 2016). They are able to provide a more holistic
understanding of the subject and enhance description and explanation. In many cases, there is a core
component, that is supplemented with an additional component to improve the depth of the research
findings. Using multiple methods to examine the same subject, can also increase the validity of the
results (Hurmerinta-Peltomaki & Nummela, 2006). Even though mixed methods provide a number of
benefits, its popularity is limited by the fact that its implementation requires a large amount of
resources, including time, financial means and skills (McKim, 2017).
Using a combination of quantitative and qualitative elements is frequently used in subject areas such
as psychology, sociology and education. It is less common in the field of transportation, which is still
dominated by quantitative-only studies. However, since the pivotal reviews of Grosvenor (2000) and
Clifton and Handy (2003), qualitative research, and as a result mixed methods, have increasingly
1
Mixed methods have been seen since the 1950s, but they are formally used since the 1980s (Creswell and Plano Clark,
2007)

Matyas and Kamargianni
6
6
gained traction (Aicart, et al., 2016). Qualitative elements can be used before the quantitative elements
to examine how best to design the qualitative surveys; or after, to better explain the results that surface
from the analysis (Hesse-Biber, 2010). Qualitative elements can bring a number of benefits to
traditional qualitative surveys in the field of transportation. Grosvenor (2000) points out, that
quantitative and qualitative research should not be viewed as substitutes, but rather as complements
that can add to the understanding of travel behavior. Also, qualitative methods can be used to improve
design and extend interpretation of quantitative surveys (Clifton & Handy, 2003). Looking at more
recent viewpoints, mixed methods have been recognized as having the most potential to provide a
well-rounded understanding of choices and behavior (Clifton, 2013; Carrasco & Lucas, 2015).
We can take a closer look at the available literature breaking it down by the samples used for the two
studies. On the one hand, mixed-methods studies can be conducted with the respondents to the
qualitative and quantitative elements being different people (e.g. Baslington, 2008; Aarhaug &
Elvebakk, 2015; Karndacharuk et al., 2016). These studies use various population groups to examine
the same topic from different viewpoints and with the two methods. For example, Baslington (2008)
carries out a travel diary with children about their route choices to school and then conducts a follow-
up interview with parents in the same areas. In a more recent study, Karndacharuk et al. (2016) carry
out a survey regarding shared street space with residents, while also conducting interviews with
transportation experts. On the other hand, it is also possible, that the same participants are used for
both the quantitative and qualitative elements, this way allowing for the follow-up questions to help
understand the reasons behind the responses to the quantitative questions. A widely cited study by
Handy and Clifton (2001) carry out a mail out mail back household travel survey about travel mode
choice for shopping trips. Following the survey, they use a selection of respondents and through focus
groups further explore the factors, motivations and attitudes behind the choices seen in the survey
responses. Other studies use in-depth interviews with a sample of survey respondents to gain deeper
insights into the outcomes of the quantitative results (e.g. Schneider, 2011; Pooley et al., 2013).
Looking more specifically at the qualitative elements, Aicart et al. (2016) provide a comprehensive
overview of recent qualitative studies in the field of travel behavior. They find that in-depth interviews
are the most common method of qualitative data collection, while focus groups come second. The two
methods are ideal in different situations. Focus groups should be used when the researcher is interested
in understanding the interactions between individuals (such as empathy or disagreement) or when the
group setting may bring out more insights (Grosvenor, 2000; Lazar, et al., 2017). However, individuals
may be less keen to share their views when others are present, especially when they may contradict
the general tendency of the group. Personal interviews remove the normative pressures and allows for
flexible types of information to be collected (Clifton & Handy, 2003). In their review, Aicart et al.
(2016) also examine the analysis techniques that studies employ when dealing with qualitative data.
They find that the most frequently applied method is thematic analysis, in which the data is explored,
to identify, analyze, organize and describe the themes and patterns that emerge (Braun & Clark, 2006;
Nowell, et al., 2017). The second most common approach is using Grounded Theory (Glaser & Strauss,
1967), which uses emerging patterns in data to generate theories, followed by the third, case studies,
which studies a person or group over time.
The above reviewed literature demonstrates that there is significant value of conducting a quantitative-
qualitative mixed methods study. So far, there is only one MaaS paper that fully incorporates both
quantitative and qualitative elements. However, this paper is based on data from a trial (where
attributes and levels are fixed to the ones tested in the trial) and their analysis is fairly limited. This
paper adds to the mixed methods literature on travel behavior by focusing on a previously little
explored topic. Second, it adds to the literature about MaaS, by examining demand using mixed
methods with data from an experiment where different product levels can be tested.

Matyas and Kamargianni
7
7
3 Methodological Overview and Data
3.1 Study Overview
The type of mixed method used in this study is an explanatory sequential design, in which the
quantitative phase is carried out first, followed by the qualitative phase to help explain the quantitative
results (Creswell, 2015). This method is ideal for research with a quantitative focus, where the
interpretation of the statistical results need additional refinements. By using qualitative methods to
follow up quantitative results, participant views on critical areas can be explored in more depth.
Following the guidelines presented in Ivankova et al. (2006) and Creswell (2015), the visual model of
procedures followed in this study is illustrated in Figure 1. The model portrays the research activities
in chronological order, alongside the products (specific outcomes) resulting from each step. In the
figure, boxes are used for data collection and analysis elements, while ovals indicate interpretation
sections, where the quantitative and qualitative phases connect with each other. QUANTITATIVE is
capitalized to indicate its priority.
Figure 1: Visual model of procedures for explanatory sequential mixed method design
In the first, quantitative phase, a web-based survey is used to collect data from 1,138
individuals. The survey consists of a revealed preference (RP) section inquiring about the
socioeconomic and current travel habits of the respondent and a stated preference (SP) experiment
specifically designed to examine preferences for MaaS plans (details to be discussed below in section
3.2). It is important to note, that during the design of the survey, a small qualitative element was also
carried out to evaluate respondents’ understanding of the survey tool. In the second step, this data is
analyzed using basic descriptive statistics and discrete choice models, namely mixed logit models.
Using the findings from the quantitative results, elements that need further, in-depth exploration are
identified in step 3. This is one of the points of inference for mixing. Interview questions are developed
focusing on key topics that are uncovered in step 2 and the interview protocol is finalized (section 5.1).
Next, qualitative data is collected through in-depth interviews with 30 people. Unfortunately, we do
not have access to the sample used for the quantitative data collection, as such, a new sample is used.
In step 5, the qualitative data is transcribed, and analyzed. Finally, step 6 integrates the quantitative
and qualitative results and inferences are drawn about how the qualitative outcomes help explain the
quantitative results.

Matyas and Kamargianni
8
8
3.2 Quantitative Survey Design and Data Collection
The survey used in this study consists of a revealed preference (RP) section inquiring about the
socioeconomic and current travel habits of the respondent and a stated preference (SP) experiment
designed to examine preferences for MaaS plans (for more information about the survey design see:
(Matyas & Kamargianni, 2017). Focusing on the MaaS SP, respondents were first presented with a
description of the MaaS concept. The text was kept short, to the point and used familiar terms to aid
comprehension and ensure that respondents actually read it. Next, 4 pages of hypothetical MaaS plans
were presented, where the respondents were asked to choose their most preferred one. Each scenario
has 4 alternatives: three fixed plans and one customizable / menu option where the users can determine
which and how much of each mode they would like. These are presented alongside each other, but
only one of them can be chosen. Thus, the outcome of a choice made from the options is either one of
the three plans or any combination of the individual features in the menu option. An example of a
scenario is presented in Figure 2. For the transport modes, icons were used with hover over
explanations. Using pictorial representations makes users’ perceptions of modes more homogenous,
makes the task more interesting and easily understandable (Morikawa, 1989).
Figure 2: Maas plans presented in SP experiment
The core attributes in the plans are the transport modes: public transport (with two levels: unlimited
bus, unlimited PT to match with the existing monthly pass options offered in London), bike sharing
(yes, no), car sharing (with levels denominated in time), and taxi (with levels denominated in distance).
Further attributes include the cost of the plan and additional special features such as ‘transferability of
unused travel to the next month’ and ’10-minute taxi guarantee’. The SP is context-dependent,
meaning that the attributes and levels shown to each respondent are dependent on our prior knowledge
about the individual. For example, if the respondent indicated in the RP element of the questionnaire
that they have disabilities and are not able to cycle, the bike sharing option is not included in their
plans. Further, a pivot design is used, whereby the SP levels are based on information from the travel
habit questionnaire that precedes the SP.
Table 1: MaaS SP attributes and levels
Attribute levels for fixed plan
Attribute levels for menu
Mode-specific attributes

Matyas and Kamargianni
9
9
Modes
Public transport
Unlimited bus
Unlimited public transport in your zones
Unlimited bus
Unlimited public
transport in your zones
None
Bike sharing
Unlimited access + 30 min use
None
Unlimited access + 30
min use None
Taxi
If current taxi usage >10 miles ->
None, current taxi usage*0.8, 1, 1.1, 1.3, 1.5
0-200 miles in
increments of 2 miles
If current taxi usage < 10 miles ->
None, 5, 8, 10, 12, 15 miles
Car sharing
If current car sharing time >0 ->
None, current car sharing time*0.8, 1, 1.1,
1.3, 1.5
0-20 hours in increments
of 1 hours;
0-7 days in increments
of 1 day
If current car sharing time = 0 ->
None, 1 hour, 2 hours, 4 hours, 6 hours
1 day, 2 days, 3 days + 2 hours
Features
Bike sharing rental time up to 60
minutes at a time
Yes, No
NA
Floating car sharing anywhere in
London
Yes, No
Car sharing includes minivan access
Yes, No
Can use any back to base car sharing
company in London
Yes, No
Pay for car sharing by the minute
Yes, No
Add an additional driver to car club
plan
Yes, No
10-minute cab guarantee
Yes, No
Only luxury cabs
Yes, No
Pooled taxi an option
Yes, No
Non-mode specific attributes
Cost
(Sum of base prices) * 0.5, 0.6, 0.65, 0.75,
0.8, 0.9, 1, 1.1
Transferability
None of your credits can be transferred to
the next month
NA
All of your credits can be transferred to
the next month
Special Present
None
NA
Free luxury car for a weekend
Free grocery delivery for a month
Free food delivery for a month
Free dinner for two
Free dinner for two (up to the value of £50)
The data collection was conducted in Greater London between November 2016 and April 2017. The
survey was dubbed the London Mobility Survey (LMS). The sample used is that of 4,558 SP
observations collected from 1,138 individuals. The characteristics of the sample are presented in Table
2, alongside the 2011 Census data and information available from the Greater London Authority (ONS,
2011; GLA, 2015). Regarding age, gender and income, LMS aligns fairly closely to the Census data.
However, ethnic minorities are significantly underrepresented as are those without household vehicles.
This, together with the limited sample size mean that the results of the study should not be generalized
to the wider population of London.

Matyas and Kamargianni
10
10
Table 2: Characteristics of sample
Variable
LMS
N=1338
Census /GLA
N=3,266,173
Gender
Male
48%
51%
Female
52%
49%
Age
18-29
23%
24%
30-39
22%
24%
40-49
17%
19%
50-
38%
35%
Ethnicity
White
83%
60%
Black
2%
12%
Asian
8%
18%
Mixed /other
7%
9%
Household Income
Up to £19,999
20%
22%
£20,000-£35,000
22%
30%
£35,000-£50,000
17%
21%
£50,000-£75,000
16%
17%
£75,000-£99,000
9%
6%
£100,000 or more
7%
4%
Prefer not to say
9%
-
Household vehicle
Has household vehicle
72%
58%
Does not have household vehicle
28%
42%
4 Quantitative Analysis and Results
4.1 Method of Analysis
One of the complexities of model development using menu data, stems from the choice set. Taking
into account all the possible combinations available to the respondent, the set of alternatives is over
26,000. Large choice sets arise from many decision contexts, such as residential location choice, trip
destination choice, route choice, vehicle acquisition choice and departure time choice. There are three
main strategies to navigate around such large choice sets: aggregation of alternatives, full sample
enumeration and sampling of alternatives. Due to the relative ease of application, in this paper we opt
to build aggregate categories of alternatives to greatly reduce dimensionality (Ben-Akiva & Lerman,
1982). This is especially useful when looking at choice sets where there are common characteristics in
alternatives (e.g. Kim et al., 2003; Carod and Antolin, 2004; Pinjari et al., 2008). Also, as a result of
the characteristics of our data, it is fairly straightforward to identify categories of alternatives. The
alternatives from the menu are aggregated based on the types of modes included in them. As there are
4 modes, public transport, bike sharing, car sharing and taxi, hypothetically there would be 42=16
possible combinations. However, the experiment constrained choices so that the respondent needed to
choose at least 2 modes, and since some combinations were not chosen, only 10 categories remained.
This, together with the three fixed plans resulted in a choice set of 13 alternatives. The resulting choice
set in the following:
Table 3: Choice set details
1
planA
2
planB
3
planC
4
menuTC
Taxi + Car Sharing

Matyas and Kamargianni
11
11
5
menuBT
Bike Sharing + Taxi
6
menuBTC
Bike Sharing + Taxi + Car Sharing
7
menuPC
Public Transport + Car Sharing
8
menuPT
Public Transport + Taxi
9
menuPTC
Public Transport + Taxi + Car Sharing
10
menuPB
Public Transport + Bike Sharing
11
menuPBC
Public Transport + Bike Sharing + Car Sharing
12
menuPBT
Public Transport + Bike Sharing + Taxi
13
menuPBTC
Public Transport + Bike Sharing + Taxi + Car Sharing
Turning to the model specification, with the aggregated choice set, the data is then used to develop
discrete choice models belonging to the family of random utility models (Domencich & McFadden,
1975; Ortuzar & Willumsen, 2001). To start off, a base multinomial logit modal(MNL) is estimated.
Due to its computational ease (closed-form mathematical structure), the significance of a large number
of parameters can be tested. To include heterogeneity, systematic taste variations are included via
interactions between the attributes of the MaaS plans and socioeconomic variables. Even though the
MNL model is the workhorse of choice modelling, it rests on a number of important assumptions
(Bhat, et al., 2016). As such, in the current application we turn to logit mixtures (mixed logit models,
MMNL), which have a much more flexible structure (McFadden & Train, 2000, Walker et al., 2007).
Most importantly for this study, mixed models are able to account for within respondent correlation
across repeated choice observations (intra-respondent taste homogeneity) (Hess & Rose, 2007). By
allowing tastes to be constant across replications for the same respondent, this panel effect can be
accounted for, which is key factor when working with SP data. Further, MMNL models allow error
components of different alternatives to be correlated, resulting in more desired structures (Blimer &
Rose, 2010).
The model used in this study is a mixed logit model with error components specification, which
captures inter-alternative correlation and panel effects. Since the random error is fixed, meaning that
it does not vary according to the observed characteristic of the person or attributes of the alternative,
the model is a normal error component logit mixture (NCLEM) (Walker et al., 2007). The error-
components structure partitions the stochastic component additively into two parts: one that allows the
unobserved error to be non-identical and non-independently across alternatives (correlated over
alternatives and heteroskedastic) and the other part to be IID over alternatives and individuals (Gumbel
distributed) (Bhat et al., 2016, Kamargianni, 2015). The former captures the correlation among
alternatives (nested structure) and correlation across responses from the same individual (panel effect)
and is distributed IID Normal (0,1) across individuals’ n but remains constant within responses t from
a given individual (Train, 2003; Walker et al., 2006).
In our model, the deterministic utility contains the attributes of the choice experiment and the socio-
economic characteristics of the participants. Prior to modelling, the key factors that were hypothesized
to affect utility were MaaS plan characteristics, individual socio-demographic characteristics and the
individual’s current mobility habits. A wide array of variables were tested including, but not limited
to, age, gender, household income, education, employment, marital status, licenses, vehicle and bicycle
ownership, awareness of shared modes, frequency of use of modes and travel pass ownership. The
final choice of variables was determined based on data availability and estimation results (significance
of t-tests of the estimated coefficients). Alternative specific constants were not estimated as the names
of the plans (Plan A, Plan B etc.) do not carry any meaning. A number of other model structures
(including nested logit, and models with additional error components) were also explored. The chosen
MMNL model was selected on the basis of statistical goodness-of-fit (likelihood ratio test, significance
of t-tests of the estimated coefficients, rho-square and adjusted rho-square).

Matyas and Kamargianni
12
12
4.2 Results
This section presents and discusses the MMNL model estimation results. The model was estimated
using the Pythonbiogeme v2.6 software (Bierlaire 2016). The number of draws was set to 500.
Table 4 provides the detailed results to the final model. The table is structured as follows: In section
1, those variables that entered the model via the fixed plans are presented. The socio-demographic
characteristics of respondents displayed subsection 1.1 are used to investigate whether there are any
significant differences in preferences for fixed plans versus menus based on the socio-demographic
characteristics of respondents. As these variables were entered into all fixed plans and none of the
menus, the menu in general provides the baseline for interpretation. Section 1.2 indicates respondents’
preferences for each of the elements within fixed plans. Even though the SP experiment had three fixed
plans presented to each respondent, the coefficients of each plan attribute are generic, as the plans
themselves (plan A, plan B, plan C) do not carry any meaning. In section 1.3 systematic taste variations
are shown through the interaction terms between the attributes of the fixed MaaS plans and
socioeconomic variables. The last two sections of the fixed plans (1.4 and 1.5) are related to the
additional features presented in fixed plans. Moving on to section 2, the variables presented in this part
are those that entered the model via the flexible plans. Sub-section 2.1 are characteristics that entered
into more than one of the flexible plan utility functions. Sections 2.2 - 2.11 present the coefficients for
variables that are specific to each of the flexible MaaS plan alternatives. Finally, section 3 provides
the error components of the model, that capture the correlation among the alternatives (nesting
structure) and the correlation across responses from the same individual (panel effect). As the number
of coefficients estimated in the model is quite extensive, in this paper we will only discuss selected
ones. We encourage the interested reader to examine Table 4 for more information about the effect of
the other factors that proved significant in the model.
In line with the research questions of this paper, we will focus on the modes and features that
respondents favored or disliked in their plans and individual characteristics that proved to be important
when choosing between MaaS plans.
Table 4: Model estimation results
Variable
Coef.
t-stat
1. Fixed Plans
1.1Socio-demographic characteristics specific to fixed plan utilities
Under 30 years of age (dummy)
5.46
4.67
In full employment (dummy)
4.67
6.32
Household income per member under £10,000 (dummy)
1.68
2.09
Under 30 interacted with full time employment
-5.49
-3.52
1.2Fixed plan specific characteristics – cost and modes
Plan cost (scaled)
-0.20
-4.44
Travelcard (dummy)
0.47
7.54
Bus pass (dummy)
0.62
8.83
Bike sharing (dummy)
-0.38
-6.01
Car sharing (dummy)
-0.40
-4.82
Car sharing days (continuous)
-0.52
-15.33
Car sharing hours (continuous)
-0.03
-3.14
Taxi (continuous)
-0.06
-12.15
1.3Fixed plan mode specific interaction terms
Under 30 (dummy) interacted with bike sharing (dummy)
0.29
1.77
Over 65 (dummy) interacted with bike sharing (dummy)
-0.25
-1.95
Bike sharing user (dummy) interacted with bike sharing (dummy)
0.33
2.86
Household owns bicycle (dummy) interacted with bike sharing (dummy)
0.29
2.79
Travelcard (dummy) interacted with travelcard ownership (dummy)
0.32
2.92
Frequent taxi user (dummy) interacted with taxi (continuous)
0.05
5.05
1.4Additional features

Matyas and Kamargianni
13
13
Additional feature(s) presented (dummy)
-0.22
-4.17
Floating car sharing included (dummy)
-0.16
-1.71
Any car sharing company can be used (dummy)
-0.38
-2.80
Car sharing can be paid by the minute (dummy)
-0.24
-1.87
Bike sharing up to 60 minutes (dummy)
-0.16
-1.72
Luxury taxis only (dummy)
-0.41
-5.12
10-minute taxi guarantee (dummy)
-0.28
-3.58
Taxipool included (dummy)
-0.53
-8.52
1.5Fixed plan additional feature interaction terms
Car sharing awareness (dummy) interacted with ‘any car sharing company can be used’
(dummy)
0.35
1.97
Under 30 (dummy) interacted with ‘bike sharing up to 60 minutes’ (dummy)
0.29
1.77
2.Flexible plans
2.1 Generic variables
Plan cost
-1.26
-7.12
Retired (dummy) specific to those plans without public transport chosen
-3.13
-4.85
Household income under £25,000 (dummy) specific to those plans where public transport
was chosen
-2.43
-3.24
Household has children (dummy) specific to those plans where public transport was
chosen
-1.77
-3.33
Travelcard was chosen (dummy) specific to those plans where public transport was
chosen
8.68
12.79
‘Taxi under 10 miles distance was chosen’ (dummy) specific to those plans where taxi
was chosen
2.13
4.75
‘Car sharing day and hour over 2 was chosen’ (dummy) specific to those plans where car
sharing was chosen
14.90
9.69
2.2 MenuTC [Taxi + Car Sharing] specific variables
Household vehicle driver (dummy)
3.21
6.10
Bike sharing user (dummy)
-2.94
-2.82
Income per household member under £10,000 interacted with under 30
4.11
3.13
2.3 MenuBT [Bike Sharing + Taxi] specific variables
Full license owner (dummy)
-6.02
-4.38
2.4 MenuBTC [Bike Sharing + Taxi + Car Sharing] specific variables
Full license owner (dummy)
-5.16
-5.79
Frequent taxi user (dummy)
2.61
2.18
2.5 MenuPC [Public Transport + Car Sharing] specific variables
Full license owner interacted with under 30
3.62
2.57
Full time employment (dummy)
-1.55
-2.23
2.6 MenuPT [Public Transport + Taxi] specific variables
Female (dummy)
2.00
4.43
Retired (dummy)
2.22
3.33
Bike sharing user (dummy)
-2.48
-2.76
Frequent taxi user (dummy)
2.91
3.07
2.7 MenuPTC [Public Transport + Taxi + Car Sharing]
Full employment (dummy)
-1.98
-3.61
Full license owner (dummy)
2.04
4.23
Bike sharing user (dummy)
-2.59
-2.65
Self-employed (dummy)
-3.31
-2.62
Household income per member over £40,000 (dummy) interacted with over 65 (dummy)
2.62
2.55
License owner (dummy) interacted with under 30 (dummy)
-3.96
-4.30
2.8 MenuPB [Public Transport + Bike Sharing]
Full license owner (dummy)
-2.37
-3.93
Household child (dummy)
2.49
2.60
2.9 MenuPBC [Public Transport + Bike Sharing + Car Sharing]
Female (dummy)
-4.81
-3.22
Full employment (dummy)
-6.08
-4.60
2.10 MenuPBT [Public Transport + Bike Sharing + Taxi]
Full employment (dummy)
-2.57
-3.65

Matyas and Kamargianni
14
14
Household vehicle driver (dummy)
-4.30
-4.89
Frequent Uber user (dummy)
3.49
1.91
2.11 MenuPBTC [Public Transport + Bike Sharing + Taxi + Car Sharing]
Retired (dummy)
3.47
3.75
3. Error component
SIGMA Fixed plans
1.05
1.81
SIGMA Menu – bike sharing
5.26
9.89
SIGMA Menu – car sharing
3.84
7.81
SIGMA Menu – public transport
-1.89
-6.08
SIGMA Menu - taxi
-5.07
-8.30
Sample size: 4558
Initial log likelihood: -6629.24
Final log likelihood: -4469.59
Rho square: 0.326
4.2.1 Evaluation of modes and features in MaaS plans
Examining the results pertinent to the research question regarding preferences for modes and features
in the plans, we approach the two plan types independently. All the transport mode specific attributes
are significant at the 99% level. Out of all the transport mode specific characteristics, only the
coefficients for the public transport options are positive, all the others are negative. The public
transport attribute took three levels: (1) none, (2) travelcard, which is the public transport seasonal
pass of London and (3) bus pass. When either one of these are included in a fixed plan (please note,
they are mutually exclusive), this plan is more preferred than those without any public transport options
in them. These positive coefficients support the theory that public transport needs to be the backbone
of MaaS systems as it is a core part of individuals’ travel preferences. Looking at the other modes,
bike sharing (binary; part of plan-not part of plan) is negative, which means that if this is included in
the plan, people prefer it less. Car sharing entered the model via three variables, all taking negative
coefficients in the model. First, a binary variable (part of plan-not part of plan) was included in the
model to capture the overall preferences of individuals for this mode. In addition, two continuous
variables for car sharing days and car sharing hours were also included. Their negative coefficients
indicate that the more that is included of these modes in a fixed plan, the less utility people gain from
them. Taxi was only entered as a continuous variable. A dummy variable was also tested, but this did
not show statistical significance. The negative sign of the bike sharing-taxi-car sharing modes will be
revisited in following sections.
The final element of the fixed plans are the additional features. Besides the features themselves, a
dummy variable was also included in the model to examine respondents’ overall preferences for
features included in fixed plans. It is very clear, that respondents are less likely to choose MaaS plans
that include any special features in them. There can be a number of reasons behind this. First, when
one or more features are presented in the plan, it becomes more complex to comprehend, which can
be a deterring factor for many people. Second, many of the features are most likely not attractive to
the general population. An interesting insight is that the feature “transferability” (meaning that unused
credit can be transferred over to next month) proved to be insignificant in the model (thus was excluded
from the final model). To examine this further, interaction terms were included to determine whether
there are any population groups who do prefer certain additional features. The model results show, that
those who were previously aware of car sharing services (and are most likely familiar with their
business models in London) are more likely to subscribe to a plan if it allows users to use any of the
car clubs in the city.
Turning to the flexible plans we will resort to basic statistical analysis to identify the frequency of each
mode chosen as well as the most popular mode combinations. Out of the 4558 choice situations,
respondents chose the create your own option in 789 (17%) of the cases. Looking now only at those

Matyas and Kamargianni
15
15
cases where the ‘create your own plan ‘was chosen; 78% of the plans that were created included some
form of public transport, 24% of plans included bike sharing, 83% included taxi, 51% included car
sharing hour and 32% included car sharing day. Looking at the combinations chosen, the most popular
one was public transport + taxi with 28%, taxi + car sharing (21%) and public transport + taxi + car
sharing (20%) was not far behind. The two least popular combinations were bike sharing + taxi and
bike sharing + taxi + car sharing.
4.2.2 Individual characteristics affecting MaaS plan choice
Overall, our hypothesis that MaaS plan characteristics, individual socio-demographic characteristics
and the individual’s current mobility habits affect utility was proven correct. However, when it comes
to socio-demographic characteristics, only a few of the tested variables were statistically significant.
When deciding between fixed vs flexible plans only age, income and employment proved to be
statistically significant. The estimated model shows that individuals with annual household incomes
under £10,000 are more likely to choose fixed plans over flexible ones compared to those with higher
incomes. This finding may, at least partially, result from the fact that the menu plans were priced
proportionally higher than the fixed plans (meaning that a fixed plan would either have the same or
lower price than a menu where the individual chose the exact same modes). This shows, that
individuals with low household incomes are not willing to pay more for the option to customize their
plans, and would rather go with the less expensive, predetermined bundles, even if this means that it
does not necessarily fit with their desires.
Similar, positive, coefficients can be seen for those who are under 30 years old as well as those in full
time employment. Holding all else equal, those who are under 30 gain higher utility from choosing
fixed plans compared to their older peers; while those in full time employment prefer fixed plans more
than those in other employment statuses. This shows some interesting insights into the preference
towards fixed products over customized ones. While allowing individuals to customize their products
has frequently been used as a successful marketing technique (Da Silveira et al., 2001), researchers
have raised certain caveats with such methods. The increasingly complex configurations have led to
consumers becoming easily overwhelmed with choice and information. Referred to as ‘consumer
confusion’, this phenomenon can negatively affect decision behavior (Matzler et al., 2007). The model
results show, that younger generations and those in full time employment are more likely to be affected
by the ‘consumer confusion’ and the additional cognitive load that results from the customizable
option. Therefore, these population segments should be targeted with either fixed MaaS plans, or ones
with only very simple customizable options.
When combining age and employment status through an interaction term, the negative coefficient
indicates that people under 30 who are in full time employment, actually prefer fixed plans less than
those under 30 year olds who are in other employment statuses. This also supports the fact that those
with less disposable income, most likely including young adults without full time employment, prefer
the relatively less expensive fixed plans.
The individual characteristics that prove important when deciding on specific MaaS plan can be broken
down into two categories: socio-demographic characteristics and current mobility habits. With regards
to socio-demographic characteristics, overall, age, gender, income, employment status and household
composition were those characteristics that had a significant impact on choice. Age was a reoccurring
socio-demographic characteristic that proved to be significant in both fixed and flexible plans. When
looking at fixed plans, those under 30 prefer bike sharing in their fixed plans more than their older
peers, while those over 65 prefer it less. Looking to the flexible plans, some examples where age is
significant include: the choice of Taxi + Car sharing, where out of the respondents with lower
household incomes those that are under 30 years old prefer this choice of plan more than their older
piers; the choice of Public transport + Car sharing, where out of the respondents with driving licenses,
those who are under 30 prefer to choose this plan combination more; and the choice of Public transport

Matyas and Kamargianni
16
16
+ Taxi + Car sharing, where out of those where the household income per member is over £40,000,
those that are over 65 years old prefer this plan combination more than younger respondents.
Employment status also appeared significant in more than one of the MaaS plans, especially with
regards to flexible plan choices. For example, retired respondents were less likely to choose plans
without public transport in them compare to those in other employment categories. Possible reasons
for this are that retirees get free public transport in London, have more free time and can take the
possible longer public transport route, can use public transport outside of rush hour when there is less
congestion inside and outside of the vehicles. Another socio-demographic factor that relates to public
transport is whether the household has children. Those respondents where there are dependent children
in the household are less likely to choose plans where public transport is included compared to those
that do not have children. As boarding and alighting from public transport can be a hassle with children,
especially younger ones, this result does not come as a surprise.
Turing to current mobility habits, these proved important in a number of places. Looking at a couple
of concrete examples, current travelcard ownership plays an important role when choosing whether a
plan should include public transport or not in both fixed and flexible plan choices. In the fixed plans,
when current travelcard ownership is interacted with travelcard, the coefficient is positive and
significant implying that those who own travelcards now would prefer to have these in their fixed plans
more than those who do not. Likewise, the same can be seen with the flexible option, those who own
a travelcard will more likely chose to include public transport in their plans than those that do not.
Public transport is not the only mode where current mobility behavior influences plan choice. Those
who have previously used bike sharing prefer to have this mode in their fixed plans more than those
have not used this mode. When it comes to creating your own plan, those respondents who have
previously used bike sharing are significantly less likely to create mode combinations that do not have
bike sharing in it (taxi + car sharing, public transport + taxi, public transport + taxi + car sharing).
Likewise, frequent taxi users (where ‘frequent’ was defined as using taxi at least once a week) are
more likely to choose fixed plans with taxi in them and create flexible plans with taxi in them (bike
sharing + taxi + car sharing, public transport + taxi, public transport + bike sharing + taxi) than those
who do not use taxi often. The sign and significance of these coefficients, show that the prior behavior
of the user is important when choosing MaaS plans, and that the blanket ‘not prefer certain modes’
that was discussed in section 4.1.1 is not necessarily true. This also shows that the prior behavior of
the user is very important when choosing MaaS plans and also shows habit persistence in preferences
when it comes to choosing fixed MaaS plans. This does not come as a surprise though, as it has been
widely studied that travel is a habitual behavior (Schlich and Axhausen 2003; Friedrichsmeier et al.
2013; Polydoropoulou et al. 2013).
5 Qualitative Phase
5.1 Data Collection and Analysis
After the quantitative analysis, some key areas of interest were identified. As the quantitative survey
restricted the question and answer frame (meaning that potentially critical information may not be
captured) the qualitative phase was purposefully conducted without a strict structure. As such, a semi-
structured discussion guide was created covering the selected areas, which were used during the
interview. The questions included in the interview focused around three priority areas where in-depth
qualitative insights could complement the findings from the quantitative models. The three areas are
(1) General understanding of MaaS, (2) Evaluation criteria of MaaS plans, (3) Overall opinion of the
concept. The discussion guide included one or two main questions and a number of follow-up prompts
for each area to help navigate the interview. Participants were encouraged to provide their own reasons
for their answers in their own words and if necessary were reminded that there is no right or wrong
answer. Question prompts were used selectively depending on the flow of the discussion. As the focus
of this paper is exploring preferences for MaaS plans, we will solely focus on the second priority area

Matyas and Kamargianni
17
17
in the following. The main question about this priority area and the prompts/follow-up questions are
presented in Table 5.
Table 5: Semi-structured questions and prompts used in interview
Priority area
Main question
Follow-up prompts
Evaluation criteria of
MaaS plans
What was going through your mind
when you were choosing between the
presented MaaS plans?
• Most important factors you were looking
for when considering the plans?
• Anything you were specifically looking
for and why?
• Anything that specifically
discouraged/deterred you from choosing
certain plans and why?
• Specific modes?
• Additional features?
• Price vs mode tradeoff?
We did not have access to the same respondents who completed the quantitative survey so a new
sample had to be recruited. To recruit participants a mix of convenience and purposive sampling was
used. Individuals were eligible if they lived within Greater London and were over 18. Participants with
various socio-demographic characteristics (age, gender, family status, vehicle ownership) were chosen
to ensure that different views were heard (see Table 6). It needs to be pointed out that the sample did
not include anyone in the age group 65+, however, as this is not the primary target audience of MaaS,
we believe this is not a major drawback at this stage. It should also be emphasized that the purpose of
the qualitative research is not to gain a representative sample, but to illustrate important themes that
may arise (and can be examined further in future research). 30 participants took part in the study and
the interviews took place during June-July 2018. Each interview lasted between 45 and 75 minutes,
depending on how long it took the participant to complete the survey and how interested they were
during the interview. The interviews were audio-taped (with written consent from the participants) and
transcribed.
Table 6: Characteristics of participants
Characteristic
Group
Number of
participants
Gender
Male
Female
12
18
Age
18-25
26-45
45-65
65+
8
11
11
0
Employment
Full time job
Part time job
Student
Other
13
7
7
3
Household composition
No children
Children
14
16
Household vehicle ownership
Yes
No
20
10
Household vehicle driver
Yes
No
17
13
To start the examination, the transcripts were transferred into the NVivo qualitative data analysis
software, through which a systematic approach could be taken to the analysis. Next, transcripts were
closely scrutinized and pertinent excerpts were assigned a provisional conceptual code or codes (if the
same text related to two or more separate codes). Codes are words, phrases or sentences that summarize
interesting and relevant features of the data. They help organize your data into meaningful groups

Matyas and Kamargianni
18
18
(Braun & Clark, 2006). The codes were arranged hierarchically, with wide, general groups at the top
and narrower sub-groups at the bottom. By systematically reading and rereading the transcripts
semantic themes emerged, which were refined and cultivated throughout the process.
5.2 Results
Before diving in, a few contextual points are worth mentioning. Travel habits and lifestyle contexts
varied greatly within the sample. Some were devoted public transport users, who also exhibited
multimodal travel involving cycling or taxis. Others were captive car users, who either for personal or
geographical reasons commute mainly with private vehicles. From the discussion about the evaluation
criteria of MaaS plans, some clear themes emerged, which are presented below. Quotations are
provided to illustrate themes and related concepts and an identifier is used to indicate the characteristics
of the respondent. The code represents their gender (Male or Female), age group (18-25, 26-45, 46-
65) and vehicle use (Driver or Non-driver) respectively. The interviews provided a wealth of insights,
however, in this paper we will focus only on selected elements that extend and supplement the findings
from the quantitative analysis.
5.2.1 Preference for public transport
The analysis confirmed that public transport is indeed a key factor in the preference towards specific
MaaS plans. When discussing how participants evaluated between MaaS plans, the essential role of
public transport was regularly mentioned. Almost all interview participants highlighted the critical role
public transport plays in their regular daily travel behavior and how this would be the core element of
their MaaS plans:
“I use public transport so much because I work in East London, but I live in West London. I'm
always travelling to London and through London, so that is obviously something that I would
be looking at straight away.” F18-25D
“I thought mostly about what it is that I use, so I thought cycling and public transport.” F46-
65N
“I was deciding more on what I personally feel I would use - I would only really use bus travel
and public transport.” M18-25N
The preference towards public transport in MaaS plans remained consistent regardless of whether the
respondent was a car user or not. Even people who indicated strong dependence towards using their
car for reasons such as having to carry their children to activities, the geographical distance of their
workplace or their sheer love of using their car expressed that out of all the modes in the MaaS plans,
public transport was the one that most interested them:
“If that was put into a package - unlimited public transport, then I would actually consider,
should I get the tube up today?” F46-65D
“If I still keep the commuting method, my private car, I could still use MaaS for all my private
stuff - collecting daughter, weekend trips around, visiting friends and family and there I was
really looking at the unlimited public transport and any of the other perks.”M36-45D
Another factor that illustrated the prominence of public transport is its relative importance compared
to the other transport modes. Several participants commented on the fact that even when other modes
were important, public transport was still the top priority:
“The specific modes I am looking for according to my current daily mobility pattern is
unlimited public transport, that’s the main one, then its taxi trips.”M26-45N
“The first thing I looked at was whether they had public transport and the second thing was
whether they had the bike sharing because they are the two things that I use every day at least
twice a day so they had to be in the plan really.” M18-25N

Matyas and Kamargianni
19
19
Finally, to solidify the idea that public transport should provide the backbone of MaaS (at least in the
case of London), the analysis revealed that a number of people would not even consider plans without
public transport in them:
“I would pretty much need unlimited travel on public transport network - so that one was the
top one I was looking for and would generally avoid the plans without that.” M26-45D
“Yes, if it didn't have that [public transport] in the plan, I wouldn't use it at all.”M18-25N
This means, that even if other plans provided much more at a relatively cheaper price, most people
would not consider those plans. All the other transport modes are frequently viewed more as
supplements to public transport, rather than core elements of individuals’ mobility portfolio.
5.2.2 Caveats to car share
The concept of car sharing, the role of car sharing in MaaS plans and the preference of respondents
towards this mode were topics that were prominent in every interview. Most participants were keen to
share their views on this mode without having to be prompted. Car sharing appeared to be a mode that
a number of respondents did not want to have in their MaaS plans:
“The car share stuff, I couldn't really see myself using so I didn't really place any value on
those.” M18-25N
“I saw car sharing and it was an obvious reject because I assumed it was more expensive.”
M26-45N
“I don't think I'd do car share” F46-65D
“I don't feel comfortable actually being the driver, so that eliminated any car shares. Car
sharing and bike sharing I just didn't want them.” F26-45N
The example quotes highlight a theme that may provide an explanation to the negative coefficient sign
of car sharing (and bike sharing and taxi) observed during the quantitative analysis. Through the
discussion, it emerged that many people would directly reject plans that included modes they would
not want, and would not even examine these. In these cases, disregarding any plans with these
undesired modes meant that the choice set that respondents actually considered was smaller. As a result
of the clear rejection of certain plans, in a number of cases people ended up selecting plans that cost
more for relatively less mobility. This points to something that could be taken into account during
future Maas plan modelling exercises; and that is the consideration choice set. There is body of
literature aimed at exploring the impact of consideration effect on consumer choice (Ben-Akiva &
Boccara, 1995; Basar & Bhat, 2004; Hauser, 2014), however it is out of the scope of this paper to delve
into this in more detail.
Even though a number of respondents exhibited the choice behavior pattern described above, in many
cases, respondents said that they would be interested in car sharing in theory, and they see how this
could be a great approach to people decreasing their use and ownership of cars.
“I think I would try and buy a bike and then get rid of the car and buy a car share of a certain
number of hours.” M26-45D
“The other two were a bit of an unknown because I don't use car share – although I think the
car share idea is really brilliant. I do know of them, but I think they are something that we
would use more if we were aware, or there was a discount or there was a chance to use it once
a month and make the most of it.” F46-65N
A common theme that emerged through the analyzed texts is that much more people were interested
in car sharing than actually picked it due to concerns regarding service quality and availability.
Participants expressed a number of caveats that made them steer away from this choice now, but if

Matyas and Kamargianni
20
20
they were ensured about these, car sharing could in fact be an important part of their MaaS plans.
Factors such as reliability, security, cleanliness, availability of child seats and location of vehicles were
among those that were mentioned:
“Maybe if it was something that was available in the area and it was a definite that I could
always access it, then I would definitely consider getting rid of my car and using a car share if
it had the things that I need - if I knew that there was always going to be a baby seat in the
trunk or something that I could use and then if it was close.” F26-45D
“The car sharing, since I've never used one, yes that’s something I could consider, but it would
have to have flexibility and speed of access and that it’s not too far to walk to get one.” M26-
45N
“As long as I knew it [car sharing vehicle] was reliable, safe, clean – definitely” F26-45D
“The whole car share thing for me - I like the idea of it, but I wouldn't necessarily car share,
because I have children and I would need to know that the car was clean and driven by a non-
smoker - one of my daughters has asthma so if a smoker was using the car, I wouldn't be able
to share with strangers.” F26-45D
5.2.3 Indifference to features
The analysis of the transcripts corroborated the negative coefficient signs in the model results for the
additional features presented alongside the modes in the MaaS plans. As these additional features
mainly showed up in more ‘complex’ plans which included more modes at an increased price, a
number of respondents expressed that the increase in price alongside the disinterest of the features
would drive them away from choosing those plans.
“They tended to be more expensive than I probably could see any reason to go for and
knowing that they would also be features that I probably wouldn't be using.” F46-65D
“I think I noticed a couple of them, but they were nothing to draw me in on that sort of price
per month.” M26-45D
“I did look at them, but then I did look at the price as well, so they did influence my decision,
but they didn't influence me enough to warrant me selecting that option.” M18-25N
These additional features, that were linked to certain transport modes such as ‘10-minute taxi
guarantee’ or ‘includes minivan access’, did not play an important role in the evaluation of plans. Most
participants did not even notice them as their main focus was on the modes and the price in the plans.
“I did see them. They wouldn't swing the decision for me.” M18-25N
“I noticed them, but they wouldn't probably weigh in.” F46-65D
“I didn't really notice them, I did notice the minivan one, but I didn't even notice the luxury cab
one, so it didn't make any difference. I just focused so much on the unlimited public transport
and the bikes that I didn't really pay attention to what’s in the car sharing option.” F26-45D
6 Conclusions
To bring together some of the main results from the quantitative and qualitative phase; both the
quantitative and qualitative results support the idea that public transport is a critical element of MaaS
plans. The preference towards public transport remains consistent regardless of the socio-demographic
characteristic or car ownership of the respondents. This was the mode that respondents looked for first,
and would not consider plans without this. Regarding the other modes (bike sharing, car sharing, taxi),
the quantitative results indicated that overall, people have a strong disfavor towards these as shown by
the negative coefficient signs. Looking into the reason for this through the qualitative analysis, it

Matyas and Kamargianni
21
21
became clear that many people directly reject plans that include modes they would not want and not
even examine these. This causes them to select plans that from a modelling perspective may not seem
rational. Even though these negative views seem to indicate that plans overall may not make sense,
both the quantitative and qualitative analysis provide some insights to the contrary. First, the
quantitative results also show that people who currently use these modes are more likely to favor them
in their plans due to habit persistence. Second, the quantitative analysis highlighted a theme, that many
people would indeed be interested in certain modes other than public transport if certain level of service
was guaranteed. Finally, both the quantitative and qualitative results indicated that individuals do not
want mode specific additional features in their plans.
There are a couple of limitations of this work that need to be noted. Regarding the stated preference
design, an efficient design should be used for future surveys as they provide the most reliable results.
Also, we recommend allowing for a no choice option. With regards to the qualitative element, given
the small sample, the limitations with regards to the external validity of the findings are acknowledged.
Even with these limitations, this paper provides valuable insights to the research community, the
public and private sector. As there is still much to learn about consumer demand and user preferences
for MaaS, this paper can inspire further research in the area.
7 Acknowledgement
This research is supported by the RCUK Centre of Energy Epidemiology and the EPSRC. The authors
of this paper would like to thank Sridhar Raman for assisting in the development of the survey tool.
8 References
Aarhaug, J. & Elvebakk, B. (2015) The impact of universally accessible public transport–a before
and after study. Transport Policy, 44, 143-150.
Aicart, M., del Lidón, M., Arroyo-López, M. R. & Ruiz Sánchez, T. (2016) Qualitative research in
travel behavior studie. Transportation research procedia , Volume 18, pp. 434-445.
Basar, G. & Bhat, C. (2004) A parameterized consideration set model for airport choice: an
application to the San Francisco Bay Area. Transportation Research Part B, Volume 38, pp. 889-
904.
Baslington, H. (2008) School Travel Plans: Overcoming Barriers to Implementation. Transport
Reviews, 28, 2, 239–258.
Ben-Akiva, M. & Boccara, B. (1995) Discrete choice models with latent choice sets. International
Journal of Research in Marketing, Volume 12, pp. 9-24.
Ben-Akiva, M. & Lerman, S. R. (1982) Discrete Choice Analysis: Theory and Application to Travel
Demand.. Camebridge(MA): MIT Press.
Bhat, C. R., Eluru, N. & Copperman R. B. (2016) Flexible Model Structures for Discrete Choice
Analysis. In: Handbook of Transport Modelling, pp.75-104.
Bierlaire, M. (2016) PythonBiogeme: a short introduction. Report TRANSP-OR 160706, Series on
Biogeme. Transport and Mobility Laboratory, School of Architecture, Civil and Environmental
Engineering, Ecole Polytechnique Fédérale de Lausanne, Switzerland.

Matyas and Kamargianni
22
22
Bliemer, M. C., & Rose, J. M. (2010) Construction of experimental designs for mixed logit models
allowing for correlation across choice observations. Transportation Research Part B:
Methodological, 44(6), pp. 720-734.
Braun, V. & Clark, V. (2006) Using thematic analysis in psychology. Qualitative Research in
Psychology, 3(2), pp. 77-101.
Carod, J. M. A., & Antolín, M. C. M. (2004) Firm size and geographical aggregation: an empirical
appraisal in industrial location. Small Business Economics, 22(3-4), 299-312.
Carrasco, J. A. & Lucas, K. (2015) Measuring attitudes: quantitative and qualitative methods.
Transport Research Procedia , Volume 11, pp. 165-171.
Clifton, K. (2013) Collecting Qualitative and Quantitative Data on the Social Context of Travel
Behaviour. In: M. L. M. M. a. J. C. (. J. Zmud, ed. Transport Survey Methods: Best Practice for
Decision Making. Bringley: Emerald Press, pp. 441-448.
Clifton, K. & Handy, S. (2003) Qualitative methods in travel behaviour research. Kruger Park,
South Africa, pp. 283-302.
Creswell, J. W. (2003) Research design: Qualitative, quantitative, and mixed methods approaches.
Thousand Oaks (CA): SAGE.
Creswell, J. W. (2015) A Concise Introduction to Mixed Methods Research. s.l.:Sage.
Creswell, J. W. & Plano Clark, V. L., 2007. Designing and conducting mixed methods research.
Thousand Oaks (CA): SAGE.
Cubic Transportation Systems (2016.)Mobility as a service: Moving towards demand-based travel.
Da Silveira, G., Borenstein, D., & Fogliatto, F. S. (2001) Mass customization: Literature review and
research directions. International journal of production economics, 72(1), pp. 1-13.
Datson, J., 2016. Mobility as a service: Exploring the opportunity for mobility as a service in the
UK.
Domencich, T. A., & McFadden, D. (1975) Urban travel demand-a behavioral analysis.
Ebrahimi, S., Sharmeen, F. & Meurs, H., 2018. Innovative Business Architectures (BAs) for Mobility
as a Service (MaaS) - Exploration, Assessment, and Categorization Using Operational MaaS Cases.
Washington D.C.
Friedrichsmeier, T., Matthies E. & Klöckner, C. A. (2013) Explaining stability in travel mode
choice: An empirical comparison of two concepts of habit. Transportation Research Part F: Traffic
Psychology and Behaviour, 16, pp. 1-13.
Georgakis, P. et al. (2018) Deliverable 2.3 - MaaS4EU use cases and reference architecture of the
EU H2020-funded project.
Glaser, B. & Strauss, A., 1967. The Discovery of Grounded Theory: Strategies for Qualitative
Research Chicago.
González Alonso, M., van Oort, N., Cats, O. & Hoogendoorn, S. (2017) Urban Demand Responsive
Transport in the Mobility as a Service Ecosystem: Its Role and Potential Market Share.
Greater London Authority (GLA) (2015) Household Income Estimates for Small Areas. Retrieved
from https://data.london.gov.uk
Grosvenor, T. (2000) Qualitative Research in the Transport Sector. Resource paper for the
Workshop on Qualitative/Quantitative Methods.

Matyas and Kamargianni
23
23
Guiltinan, J. (1987) The price bundling of services: A normative framework. The Journal of
Marketing, pp. 74-85.
Hauser, J. R. (2014) Consideration-set heuristics. Journal of Business Research, 67(8), pp. 1688-
1699.
Heikkilä, S. (2014) Mobility as a service-a proposal for action for the public administration, case
Helsinki'.
Hensher, D. (2017) Future bus transport contracts under a mobility as a service (MaaS) regime in the
digital age: Are they likely to change?. Transportation Research Part A, Volume 98, pp. 86-96.
Hess, S., & Rose, J. M. (2007) Intra-respondent taste heterogeneity in instantaneous panel
surveys. Arbeitsbericht Verkehrs-und Raumplanung, 448.
Hesse-Biber, S. (2010) Qualitative approaches to mixed methods practice. Qualitative Inquiry,
Volume 16, pp. 455-468.
Hietanen, S. (2016) Mobility as a Service’ – the New Transport Model?. Eurotransport, 12(2), pp.
133-176.
Ho, C., Hensher, D., Mulley, C. & Wong, Y. (2017) Prospects for switching out of conventional
transport services to mobility as a service subscription plans – A stated choice study. Stockholm,
Sweden.
Hurmerinta-Peltomaki, L. & Nummela, N. (2006) Mixed methods in international business research:
A value-added perspective. Management International Review, Volume 46, pp. 439-459.
Ivankova. N. V. Creswell, J. W. & Stick, S. L. (2006) Using Mixed-Methods Sequential Explanatory
Design: From Theory to Practice. Field Methods, 18(1), pp. 3-20.
Jittrapirom, J., Caiati, V., Feneri, A.-M. ., S. Ebrahimigharehbaghi, S., Alonso- González, M. J. and
Narayan J. (2017) Mobility as a Service: A Critical Review of Definitions, Assessments of Schemes,
and Key Challenges. Urban Planning, 2(2), pp. 13-25.
Johnson, R. B. & Onwuegbuzie, A. J. (2004) Mixed methods research: A research paradigm whose
time has come. Educational Researcher, 33(7), pp. 14-26.
Kamargianni, M., Matyas, M., Li, W. & Schafer, A. (2015) Feasibility Study for “Mobility as a
Service” Concept in London. MaaSLab - UCL Energy Institute Report, Prepared for Department for
Transport.
Kamargianni, M., Li, W, Matyas, M & Schäfer, A. (2016) A Critical Review of New Mobility
Services for Urban Transport. Transportation Research Procedia, Volume 14, pp. 3294-3303.
Kamargianni, M. & Matyas, M. (2017) The Business Ecosystem of Mobility-as-a-Service. Paper
presented at the Transportation Research Board Annual Meeting. Washington D.C.
Karlsson, A., Sochor, J. & Strömberg, H. (2016) Developing the ‘Service’ in Mobility as a Service:
experiences from a field trial of an innovative travel brokerage. Transportation Research Procedia,
Volume 14, pp. 3265-3273.
Karndacharuk, A., Wilson, D.J., & Dunn, R.C.M. (2016) Qualitative evaluation study of urban
shared spaces in New Zealand. Transportation Research Part D, 42 119-134.
Kim, J., Pagliara, F., Preston, J. (2003) An analysis of residential location choice behaviour in
Oxfordshire UK—a combined state preference approach. International Review of Public
Administration 8, 103–114.

Matyas and Kamargianni
24
24
Lazar, J., Feng, J. H. & Hochheiser, H. (2017) Research methods in human-computer interaction.
Morgan Kaufmann.
Li, Y. & Voege, T. (2017) Mobility as a Service (MaaS): Challenges of Implementation and Policy
Required.. Journal of Transportation Technologies, Volume 7, pp. 95-106.
Magoutas, M. et al. (2017) Deliverable 2.1 -State-of-the-art-report of the EU H2020-funded project
MaaS4 EU.
Matyas, M. & Kamargianni, M. (2017) Stated preference design for exploring demand for “mobility
as a service” plans. Presented at the International Choice Modelling Conference, Cape Town, South
Africa.
Matzler, K., Waiguny, M., & Fuller, J. (2007) Spoiled for choice: consumer confusion in Internet-
based mass customization. Innovative Marketing, 3(3), pp. 7-18.
McFadden, D., & Train, K. (2000) Mixed MNL models for discrete response. Journal of applied
Econometrics, 15(5), 447-470.
McKim, C. A. (2017) The value of mixed methods research: A mixed methods study. Journal of
Mixed Methods Research, 11(2), pp. 202-222.
Morse, J. M. (2016) Mixed method design: Principles and procedures. Routledge.
Morikawa, T. (1989) Incorporating stated preference data in travel demand analysis, Doctoral
Dissertation, Massachusetts Institute of Technology.
Mulley, C. (2017) Mobility as a Services (MaaS) – does it have critical mass?. Transport Reviews,
37(3), pp. 247-251.
Nowell, L. S., Norris, J. M., White, D. E. & Moules, N. J. (2017) Thematic Analysis: Striving to
Meet the Trustworthiness Criteria. International Journal of Qualitative Methods, 16(1).
Office for National Statistics (ONS) (2011) Census: Digitised Boundary Data (England and Wales)
[computer file]. UK Data Service Census Support. Downloaded from: https://
borders.ukdataservice.ac.uk/
Ortuzar, J. D. & Willumsen L.G. (2011) Modelling Transport. John Wiley & Sons.
Pinjari, A., Eluru, N., Bhat, C., Pendyala, R. & Spissu, E. (2008) Joint model of choice of residential
neighborhood and bicycle ownership: accounting for self-selection and unobserved
heterogeneity. Transportation Research Record: Journal of the Transportation Research Board,
2082, pp. 17-26.
Polydoropoulou, A., Kamargianni, M. & Tsirimpa, A. (2013) Car Use Addiction vs. Ecological
Consciousness: Which one Prevails on Mode Choice Behavior? In book: Travel Behaviour
Research (IATBR), Editors: Roorda, M., and E. Miller, pp. 128-139.
Pooley, C. G., Horton, D., Scheldeman, G., Mullen, C., Jones, T., Tight, M., Jopson, A, & Chisholm,
A. (2013) Policies for promoting walking and cycling in England: A view from the street. Transport
Policy, 27, 66-72.
Schlich, R.& Axhausen, K. W. (2004)Habitual travel behaviour: evidence from a six-week travel
diary. Transportation, 30(1), pp. 13-36.
Schneider, R.J. (2011) Understanding Sustainable Transportation Choices: Shifting Routine
Automobile Travel to Walking and Bicycling.

Matyas and Kamargianni
25
25
Smith, G., Sochor, J. & Karlsson, I. M. (2017a) Mobility as a Service: Implications for future
mainstream public transport. Presened at Thredbo15 - International Conference Series on
Competition and Ownership in Land Passenger Transport, Stockholm, Sweden.
Smits, G., Sochor, J. & Karlsson, M. A. (2017b) Procuring Mobility as a Service: Exploring
dialogues with potential bidders in West Sweden. Presented at ITS World Congress, Montreal.
Sochor, J., Arby, H. & Karlsson, M. A. (2017) The topology of Mobility as a Service: A tool for
understanding effects on business and society, user behavior, and technical requirements. Montreal.
Sochor, J., Strömberg, H. & Karlsson, M. A. (2015) Implementing Mobility as a Service: Challenges
in Integrating User, Commercial, and Societal Perspectives. Transportation Research Record:
Journal of the Transportation Research Board, Volume 2536, pp. 1-9.
Tashakkori, A. & Teddlie, C. (2010) Sage handbook of mixed methods in social & behavioral
research. Sage.
Walker, J. L., Ben-Akiva, M., & Bolduc, D. (2007) Identification of parameters in normal error
component logit-mixture (NECLM) models. Journal of Applied Econometrics, 22(6), pp. 1095-1125.
Wong, Y. Z., Hensher, D. A. & Mulley, C. (201). Emerging transport technologies and the modal
efficiency framework: A case for mobility as a service (MaaS). Presened at Thredbo15 -
International Conference Series on Competition and Ownership in Land Passenger Transport,
Stockholm, Sweden.