Factors aﬀecting the adoption and use of urban air mobility
Christelle Al Haddada,∗
, Emmanouil Chaniotakisb, Anna Straubingerc, Kay Pl¨otnerc, Constantinos
aChair of Transportation Systems Engineering, Technical University of Munich, Arcisstrasse 21, 80333 Munich, Germany
bMaaSLab, Energy Institute, University College London, 14 Upper Woburn Place, WC1H0NN London, UK
cBauhaus Luftfahrt e.V., Willy–Messerschmitt–Strasse 1, 82024 Taufkirchen, Germany
Technological advances have recently led to the development of urban air mobility (UAM), an alternative
transportation mode with several concepts including vehicles operated by on-demand fully-automated verti-
cal take-oﬀ and landing aircraft (VTOL) for intra–city passenger transportation. However, despite a growing
interest in UAM, understanding users’ perceptions to it remains limited. This research aims to identify and
quantify the factors aﬀecting the adoption and use of UAM, based on relevant tools from the literature,
such as recurring factors in studies on aerial vehicle concepts, ground autonomous vehicles, but also accep-
tance models, such as the Technology Acceptance Model by Davis et al. (1989). A stated–preference survey
was developed to assess the perception of users in terms of adoption time horizon, including options such
as the ﬁrst six years of the service’s implementation, “unsure”, and “never”. The obtained results were
evaluated using exploratory factor analyses, and the speciﬁcation and estimation of suitable discrete choice
models, multinomial logit models (MNLs) and ordered logit models (OLMs), with adoption time horizon
as dependent variable. Findings revealed the importance of safety and trust, aﬃnity to automation, data
concerns, social attitude, and socio–demographics for adoption. Factors, such as the value of time savings,
the perception of automation costs, and service reliability, were also found to be highly inﬂuential. There
was also an indication that skeptical respondents, i.e. answering “unsure”, had a behavior similar to late
and non-adopters, i.e. adoption time horizon higher than six years or answering “never”. The summarized
results were represented in an extended Technology Acceptance Model for urban air mobility, and provided
insights for policymakers and industrial stakeholders.
Keywords: urban air mobility, adoption, acceptance, technology acceptance model, perception, stated
preference, exploratory factor analysis, discrete choice modeling
Shared mobility services are providing users with more eﬃcient travel options, characterized by lower
demand for parking spaces, lower vehicle ownership, but also reduced environmental impacts resulting from
lower emissions (Baptista et al.,2014). At the same time, autonomous vehicles promise safe and comfortable
transportation, and most automobile manufacturers are likely to launch fully autonomous vehicles in the5
coming decade (Bimbraw,2015). This trend has led to a research interest in ground shared autonomous
mobility (Fagnant & Kockelman,2014), and possibly to the exploration of the third dimension: the skyscape.
This has been reﬂected in aerial vehicle concepts for passenger transportation, also called as “urban air
mobility” (UAM). The concepts for short–haul passenger air trips are facilitated by technological advances
Email addresses: firstname.lastname@example.org (Christelle Al Haddad), email@example.com (Emmanouil
Chaniotakis), Anna.Straubinger@bauhaus-luftfahrt.net (Anna Straubinger), Kay.Ploetner@bauhaus-luftfahrt.net (Kay
Pl¨otner), firstname.lastname@example.org (Constantinos Antoniou)
Preprint submitted to Transportation Research Part A; https: // doi. org/ 10. 1016/ j. tra. 2019. 12. 020 January 2, 2020
in terms of battery storage, electrical power transmission and distributed propulsion systems (Shamiyeh10
et al.,2017). The US National Aeronautics and Space Administration (NASA) is developing a framework for
the integration of urban air mobility airspace research with the eﬀorts of diﬀerent partners and stakeholders
(Thipphavong et al.,2018). For instance, Airbus (2018) introduces urban air mobility as the integration of
the third dimension to urban transport networks, including the on–demand sharing mobility operated by
vertical take–oﬀ and landing aircraft (VTOL) for intra–city passenger trips, with a long–term vision entailing15
electrical self–piloted VTOLs, such as Vahana and CityAirbus demonstrators, for one or more passengers
respectively. The introduction of urban air mobility, according to Airbus (2017), promises to oﬀer a safer,
more reliable, and more environmental alternative to alleviate congestion on transport networks. Currently,
Voom (an Airbus company) is providing on–demand helicopter bookings in megacities like Mexico City and
S˜ao Paulo, laying the ground for further developments in urban air mobility (Airbus,2018). Uber has also20
introduced air taxi models in their business plans, such as the economic model assuming a four–seat capacity
(including the pilot if the vehicle is not self–piloted), for which passengers have the possibility to share the
ﬂight and thereby save on the ride cost (Uber Elevate,2016). Porsche Consulting (2018) deﬁnes the process
of this service as follows: starting from the ﬁrst mile (access to the vertiport), followed by the boarding
on the vehicle, the ﬂight time (including take–oﬀ and landing), the de–boarding, and ﬁnally the last mile25
transfer; the service can be booked online and the assumed boarding and de–boarding times are around
three minutes each. In this study, Porsche Consulting (2018) presents three e–VTOL concepts with diﬀerent
certiﬁcation times, travel speeds (from 70 to 300 km/h), routes, and purposes (intra– and/or intercity trips),
projecting for 2035 a passenger market of around 23 000 aircraft and worth $32 Billion.
Despite diﬀerent model deﬁnitions, urban air mobility is constrained by many aspects related to regula-30
tions, infrastructure availability, air traﬃc control, environmental impacts, but also community acceptance
(Vascik,2017). In terms of infrastructure challenges, Cohen (1996) studied vertiport prototypes to optimize
land use, site selection, and community acceptance and more recently Fadhil (2018) focused on a GIS–based
analysis for ground infrastructure selection for urban air mobility. In modeling UAM impacts, the service
implications on the inhabitants and the changes on the city were evaluated (Straubinger & Verhoef,2018),35
but also an agent–based simulation and an initial analysis of the service’s performance were conducted
(Rothfeld et al.,2018).
To better understand the service demand and to be able to predict it, understanding consumer adoption
is crucial; in other words, if and when users are going to use urban air mobility. For instance, a study from
the Georgia Institute of Technology designed a survey to collect responses from 2,500 high–income workers40
in diﬀerent areas of the US to predict their demand for eVTOL in urban areas, with ﬂights ranging up to
90 km (Garrow et al.,2018). Passenger adoption in a UAM environment was also explored using choice
modeling in a case study based in Munich (Fu et al.,2019). Setting the context for the characteristics that
imply the use of VTOL, Antcliﬀ et al. (2016) proposes Silicon Valley as an ideal region for early introduction
of on–demand civil VTOL operations, mostly due to good weather, high housing prices, high incomes, but45
also number of hyper commuters (those who commute two or more hours per day) who would greatly reduce
their time in congestion (Antcliﬀ et al.,2016). In Germany, research on UAM has targeted the city of
Ingolstadt as part of the “Urban Air Mobility Initiative”, supported by the European Commission.
Vascik & Hansman (2018) emphasized the signiﬁcance of noise and altitude levels as areas of concerns
for community acceptance of urban air mobility systems. To the best of the authors’ knowledge, however,50
to the service characteristics) outside of a mode choice context, and in relation to adoption time horizon.
This paper helps to ﬁll this gap by developing a framework for UAM adoption and use, assuming an on-
demand intra–city passenger transportation, operated by fully-automated electric VTOL, integrated with
public transportation systems, with the possibility to ride-pool (share the ride with other passengers to55
save money). This is based on the development of a novel stated preference survey design with adoption
time horizon as a dependent variable, aiming to reveal factors aﬀecting adoption of this service. Futuristic
scenarios were introduced to evaluate a hypothetical system starting in 2030, aiming to better understand
if and when users would adopt them, but also the barriers and opportunities from a behavioral perspective.
Accordingly, factors were identiﬁed by exploring those found to be signiﬁcant in the perception of aerial60
vehicle concepts, in technology and automation acceptance models, and aﬀecting automation acceptance in
transport. This research contributes to the existing body of literature on the acceptance and adoption of
new forms of mobility by presenting a novel methodological approach using a non–conventional dependent
variable for ordered logit models: time variable in a scale including a non–ordinal parameter, such as
“unsure”, as to the best of the authors’ knowledge these models are mostly used with ordinal scales, such as65
Likert scales, for attitudes and satisfaction. The obtained results are represented in a Technology Acceptance
Model for urban air mobility that could be generalized for disruptive transport technologies. The signiﬁcantly
identiﬁed factors oﬀer a good basis for the investigation of more realistic scenarios in a nearer future, as
they show an aptitude of users towards them.
The remainder of this paper is structured as follows. A literature review is ﬁrst presented, followed70
by the methods used including the survey design, descriptive analysis and model development, and study
limitations. The data collected and main statistical results are then derived. After that, the developed and
estimated models are shown, namely the exploratory factor analyses and discrete choice models. Then, a
discussion of the main survey ﬁndings is presented along with the proposed Technology Acceptance Model
for Disruptive Transport Technologies, and the policy level insights and recommendations. Industrial stake-75
holders could thereby appropriately care for the needs of diﬀerent classes of users, while policymakers could
beneﬁt from these factors by setting guidelines or more stringent regulations if required. Finally, a conclusion
is drawn with some recommendations on future relevant work.
2. Literature review
To examine the adoption of a new service, pertinent literature suggests that it is necessary to understand80
the conditions for trusting this service and accepting its operational characteristics. In the case of UAM,
this inevitably passes through the investigation of technology and automation acceptance models, and of
factors commonly aﬀecting the acceptance of automation in transport. As only a few studies have explicitly
focused on consumer adoption of urban air mobility, examining studies on community acceptance of other
aerial vehicles or services might be useful despite its limitations.85
2.1. Factors aﬀecting aerial vehicles perception
Recent studies commissioned by NASA investigated market potentials for urban air mobility, focusing
on the diﬀerent barriers to its implementation, including community acceptance and key factors for its
users and non–users’ adoption. The study commissioned to Crown Consulting studied the viability of three
UAM use cases in 15 US cities (NASA,2019): last mile delivery (for packages), air metro (an autonomous90
public transit style commuter system) and air taxi (an autonomous on-demand ridesharing system). This
study revealed that nearly half of all surveyed consumers were potentially comfortable with delivery and
UAM use cases. Major concerns presented were safety, privacy, job security, environmental threats and
noise and visual pollution. Moreover, cybersecurity, aﬀordability and willingness to pay were perceived as
barriers to the UAM viability. The study commissioned to Booz Allen Hamilton (NASA,2018) looked at95
three potential UAM markets (airport shuttle, air taxi and air ambulance) in ten target urban areas, to
explore market size and barriers, understand the potential viability of the use cases, but also the societal
and environmental impacts of UAM; this study included a survey, focus groups, and stakeholder interviews.
Societal barriers were analyzed by looking at both user and non-user perspectives. Key factors were safety
(including passengers’ need for security screening), privacy and noise, preference for piloted aircraft, impact100
of having a ﬂight attendant, cybersecurity, cost and convenience, but also socio–demographics; men (along
with millennials and younger respondents) were found to be more comfortable and willing to use UAM.
In a stated preference survey based in Munich, Fu et al. (2019) explored transportation mode preferences
in an Urban Air Mobility environment including four alternatives: private car, public transportation, au-
tonomous taxi, and autonomous ﬂying taxi (or UAM). The study results indicated that travel time, travel105
cost, and safety might be critical for determining UAM adoption. Socio–demographics were also found as
highly inﬂuential in UAM use, as market segments showed that younger individuals as well as older ones
with higher household income are more likely to adopt UAM. Also, trip purpose proved to be signiﬁcant,
with non–commuting being the respondents’ most preferred option.
Research on consumer acceptance of unmanned civilian drones identiﬁed common concerns among people,110
such as privacy (Wang et al.,2016;Clothier et al.,2015), risks associated with accidents, and security
regarding the recognition of drones used for emergency (Lidynia et al.,2017). Additionally, Chamata (2017)
identiﬁed social and economic concerns as factors delaying the adoption of civilian drones. According to
the trip purpose, the community acceptance of drones might diﬀer (Boucher,2016). For unmanned aerial
vehicles (UAV), MacSween (2003) found that emotional and safety data increased users’ persuasions of UAV115
applications, including commercial, cargo, and passenger transportation. Whereas the above concerns apply
to community acceptance of drones, they could be very relevant to consumer adoption of UAM, as they all
discuss users’ perceptions of unmanned aviation.
A stated preference study by Peeta et al. (2008) investigating the travel propensity towards air taxi ser-
vice, deﬁned as very light jet–based on–demand air service (ODAS), predicted the probability of individuals120
of switching from their usual mode of intercity transportation to an on-demand air service (ODAS) given a
set of scenarios of travel distance and cost. The adoption of ODAS was found to be highly dependent on
travel distance, service fare, and level of accessibility. Kreimeier & Stumpf (2017) found cost to be crucial
in determining consumer adoption in Germany. The market share for thin–haul on–demand air mobility
services has been estimated as 19% or 235 million trips per year, based on a choice model assuming 0.1125
AC /km costs gap between UAM services and cars; a sensitivity analysis showed that this volume would drop
to 24 million for a gap of 0.2 AC /km. This market would be potentially viable for distances above 100 km,
with 60% of the market expected for distances between 200 and 400 km (inter–city). Both studies by Peeta
et al. (2008) and Kreimeier & Stumpf (2017) presented inter–city applications or thin–haul ﬂights, which
are arguably diﬀerent than intra–city application of UAM; however, they still highlighted cost and time as130
signiﬁcant factors for consumer adoption of on–demand aerial mobility.
2.2. Technology and automation acceptance models
Technology acceptance has long been an area of interest in research with the main purpose of modeling
how people accept and use a technology. Perhaps the most renown technology acceptance model is the
Technology Acceptance Model (TAM) by Davis et al. (1989), developed to investigate technology use of135
information systems, particularly computer technology, in which the correlation between the intention to
use and actual usage is measured. The main aim of the model is to present a framework for modeling the user
acceptance in terms of factors that inﬂuence his or her decision in using the technology. This model is based
on two main constructs: the perceived usefulness (PU) and the perceived ease of use (PEU), where PEU
reinforces PU. The former is the extent to which the user believes the technology use would enhance his or her140
job performance, whereas the latter is the degree to which using the technology requires eﬀort. Both factors
determine the user attitude towards using the system, which in turn determine the behavioral intention (BI)
to use the system, and then the actual system use. This model also includes external variables, which aﬀect
the deﬁned constructs. Based on this model, the Technology Acceptance Model 2 (TAM2) was extended by
adding variables grouped into social inﬂuence and cognitive processes (Venkatesh & Davis,2000), impacting145
the perceived usefulness of the technology. Social inﬂuence includes subjective norms, voluntariness, image;
cognitive processes included job relevance, output quality, and result demonstrability. In TAM2, these main
factors directly inﬂuence user perception (PU and PEU). A further revision extended TAM into the Uniﬁed
Theory of Acceptance and Use of Technology (UTAUT), (Venkatesh et al.,2003), and a later revision into
TAM3 (Venkatesh & Bala,2008). According to Ghazizadeh et al. (2012), the UTAUT oﬀers a more complete150
account of acceptance in comparison with TAM2. In UTAUT, three constructs directly determine BI:
performance expectancy, eﬀort expectancy, and social inﬂuence (as deﬁned in TAM2); facilitating conditions
is a fourth construct that directly determines the use behavior. Gender, age, experience, and voluntariness
moderate the impact of the four constructs. The role of moderating factors (factors impacting adoption and
reducing the limitations of the models’ explanatory power) has been found crucial (Sun & Zhang,2006) in155
technology acceptance models; for instance, the moderating eﬀects of gender and self–eﬃcacy in the context
of mobile payment adoption (Riad et al.,2014).
To combine constructs from TAM models with those from the Cognitive Engineering literature, Ghaz-
izadeh et al. (2012) proposed a framework for automation acceptance model (AAM), by adding two main
constructs: compatibility (namely task–techhnology compatibility) and trust. The former includes previous160
experience with automation, existing work practices, work style, and values; the latter (trust) inﬂuences BI
directly and indirectly through PU and PEU. The model is also reinforced by the feedback mechanisms of
the diﬀerent constructs. Zhang et al. (2019) used structural equation modeling to test hypotheses on their
proposed autonomous vehicles acceptance model, in which initial trust is the most critical factor in promot-
ing a positive attitude towards using autonomous vehicles, which along with perceived usefulness determine165
the behavioral intention (to use autonomous vehicles). In this model, initial trust is aﬀected by perceived
usefulness, perceived ease of use, but also two types of risk: the perceived safety risk and the perceived
privacy risk; initial trust can be improved by enhancing perceived usefulness or by reducing perceived safety
risk. More recently, a study at the University of St.Gallen in Switzerland developed an adapted TAM for
personal autonomous mobility, focusing on ground autonomous vehicles (Jenkins et al.,2018). The main170
additions in automation–related models could serve as a motivation for the extension or application of an
urban air mobility acceptance model.
2.3. Factors aﬀecting automation acceptance in transport
Social acceptance is a widely researched area in the ﬁeld of automation, particularly for ground au-
tonomous vehicles1. In a study on the trust and acceptance of shared autonomous vehicles (SAE2level 4),175
Merat et al. (2016) summarized the factors inﬂuencing the use of conventional ride–sharing vehicles, but also
those aﬀecting the acceptance and trust of robotic systems, where trust includes reliability and safety. Choi
& Ji (2015) also noted the importance of trust in adopting autonomous vehicles; by conducting a survey
of 552 drivers and analyzing the data using partial least squares, they noted the importance of perceived
locus of control or situation management for gaining trust. Nees (2016) found that realistic expectations,180
and thereby perceived reliability, are inﬂuential in user acceptance of autonomous vehicles. In urban areas
in particular, passenger security as part of perceived safety has been noted as crucial for gaining trust (and
thereby the implementation and adoption of AVs); especially during the night (Piao et al.,2016). The user’s
feeling of safety could depend on the vehicle’s interior (Merat et al.,2016), but might also be hindered
by cyber–security concerns (Kyriakidis et al.,2015). In Germany, 90% of the respondents of a study on185
autonomous driving stated that they would feel safer, if they were able to intervene or control the vehicle
at any time, or at least in case of emergency (Deloitte Analytics Institute,2017). Control has also been
associated with feeling more independent (Gaggi,2017) and proved to positively impact the user’s perceived
ease of use (R¨odel et al.,2014). Overall, users have to be aware and convinced of the beneﬁts of the new
technology in order to be able to trust it and therefore use it. Manufacturers’ reputation could be a positive190
factor in gaining this trust (Deloitte Analytics Institute,2017). Real–life tests for autonomous vehicles and a
higher transparency in demonstrating the new modes could also lead to a higher trust through an increased
user awareness (Bjørner,2015;Choi & Ji,2015).
A higher perception of the beneﬁts of automation could help users gain trust to use the technology, which
is reﬂected in a higher perceived usefulness. A decrease in congestion and improvement of road safety are195
associated with social beneﬁts (Kaan,2017) resulting from the reduced number of road crashes. Improved
mobility for mobility–impaired users (Clements & Kockelman,2017) can be seen as another social beneﬁt.
Moreover, the trip purpose could play a role in behavioral intention. Users might be more interested in
using autonomous vehicles for leisure trips, after alcohol consumption (Connected Automated Driving EU,
2017), or if the intended trip is unpleasant (Deloitte Analytics Institute,2017). The perceived advantages200
of automation may lead to a higher perceived usefulness, resulting in a higher user acceptance. Ease of
use in terms of eﬀort expectancy in the case of autonomous vehicles, or ease of access for dependents or
mobility–impaired users (Merat et al.,2016) might inﬂuence the intention to use autonomous vehicles. For
UAM, however, the perceived ease of use (PEU) is not related to eﬀort expectancy or driver intervention,
like in the case of ground AVs, as users are not expected at any time to take control of the air vehicle;205
instead PEU could be translated to simply the booking and boarding processes, since these are the only
tasks expected from the users.
1For practical reasons, the term “autonomous vehicles” will be used to refer to autonomous ground vehicles, and abbreviated
with AV. Otherwise, the term urban air mobility or on–demand air mobility will be used to refer to autonomous air vehicles.
2Society of Automotive Engineers
Attributes, such as time (Krueger et al.,2016) and costs (Merat et al.,2016), were found to be highly
inﬂuential in the acceptance of autonomous vehicles. Comfort (Rychel,2016), vehicle cleanliness (Merat
et al.,2016), and availability in diﬀerent weather conditions (Merat et al.,2016) proved to impact the210
intention to use; the latter may also be inﬂuenced by social behavior, including the willingness to share
the ride with strangers (Merat et al.,2016) and the perceived fun of driving or driving enjoyment (Bjørner,
2015). Social attitudes can also include concerns from the loss of jobs induced by automation (Deloitte
Analytics Institute,2017), concerns regarding terrorism, crime, or cyber–security in general, data concerns,
such as data protection, data use, or privacy in broader terms (Kyriakidis et al.,2015), but also cultural215
values since acceptance and adoption vary globally. In industrialized countries, for example, automated
vehicles might face higher skepticism compared to emerging countries like India or China (Rychel,2016).
As in the former accident rates are lower, due to higher measures of safety, people might be more reluctant
towards automation, as they are not necessarily convinced of its beneﬁts. Also, in more developed countries,
users are less likely to be comfortable with their data being shared or used (Kyriakidis et al.,2015).220
Socio–demographic factors were mentioned in several studies focusing on automation perception (R¨odel
et al.,2014;Kyriakidis et al.,2015;Payre et al.,2014). Women were often found to have a lower intention to
use autonomous vehicles, compared to their male counterparts (Hohenberger et al.,2016); in contrast, young
individuals (Deloitte Analytics Institute,2017) and individuals with multimodal travel patterns (Krueger
et al.,2016) were found to be more likely to adopt shared autonomous vehicles. Similarly, technology225
awareness was a positive impact on autonomous vehicles acceptance; this includes, for instance, having
heard of Google cars (Bansal et al.,2016), autonomous vehicles (Schoettle & Sivak,2014), and/or previous
experience with advanced driver–assistance systems (ADAS).
3.1. Survey design230
As the service in question is not yet operational and can not be observed or used, a Stated Preference (SP)
survey was designed to identify the most inﬂuential factors in UAM’s adoption time horizon. The survey
was structured in four parts, with a total of 31 questions (or question groups3). The study particularly
focused on the Munich region, and therefore the survey was available in both languages: English (EN) and
The ﬁrst and last parts of the survey targeted respondents’ travel behavior and socio–demographics,
respectively. In the ﬁrst part, respondents were asked about their commute behavior, their satisfaction with
the public transportation system of their region, and their attitudes towards automated systems (enjoyment,
trust, perceived usefulness, and previous experience with automation). As these attitudes are commonly
found in technology acceptance and use, but are rather diﬃcult to measure directly, they were assessed240
through (ﬁve–point) Likert scale agreement statements (Likert,1932), with options ranging from “strongly
disagree” to “strongly agree”; answers were verbally labelled and midpoints (“neither disagree nor agree”
used for attitude measures) were used for attitude measures, as suggested by Dolnicar (2013). These
statements included: “I enjoy interacting with automated systems, such as Siri (Apple) or Alexa (Amazon)”,
“I trust such automated systems”, “I think driver assistance systems, such as adaptive cruise control, lane245
keeping assistance, or other advanced systems, are useful”, “I have used (in my own car or someone else’s)
such advanced driver-assistance systems”. Socio–demographic questions included age, gender, household
size, disability in household, education level, main occupation, household income, and current residence
location (city and country), with a “prefer not to answer” option for these. As stated preference studies
are susceptible to anchoring bias (McFadden,2001), socio–demographic questions were placed at the end of250
the survey, so that people would not be biased in answering them according to what they think would be
consistent with their choices. This can also overcome the stereotype threat (Steele,1995) in which people
try to avoid the conﬁrmation of any stereotype they feel threatened by.
3A question consisted sometimes of a matrix including several agreement statements, focusing on one attribute for instance.
In the second part, UAM was introduced by presenting some of its properties found in the pertinent
literature (presented in Section 1). The assumptions for the vehicle design were mostly based on the economic255
model by Uber Elevate (2018) who proposed a vehicle with four seats. The process was deﬁned as in Porsche
Consulting (2018), starting with the ﬁrst mile, followed by boarding, in–vehicle ﬂight including take–oﬀ and
landing, de–boarding, and ﬁnally the last mile; in this study, both boarding and de–boarding processes were
estimated at around three minutes, with ﬂight ranges of 20 to 50 km (for intra–city). In particular, for our
study, an example of 30 km was given for the transfer from Munich’s airport to its city center (Marienplatz).260
Based on the above studies, the system was deﬁned as a future mobility service provided by fully–
automated electrically powered vertical take–oﬀ and landing aircraft (eVTOL) with four–seat capacity (in-
cluding one wheelchair seat), operating at a speed of around 150 km/h. The service would be on–demand,
booked online (through the service’s app or website), and passengers would have the possibility to book for
up to four passengers or to ride–pool with other passengers and save money. The required boarding time was265
set to be 5 minutes prior to take–oﬀ, with access and egress through vertiports (helipad–like infrastructures)
that would be integrated with the existing public transportation system.
Two scenarios were given to present realistic attributes of the service (trip duration and fare), starting
hypothetically in 2030, with taxi as a benchmark. These targeted the region of Munich, by presenting two
trips where public transportation would need more than an hour with a necessary transfer in the center,270
with intra–city urban ranges of 20 and 30 km (for the second and ﬁrst examples respectively). Other modes
were not presented as the aim of the survey was to understand the user adoption and not to conduct a mode
choice study; scenarios therefore aimed to familiarize respondents with the system. For UAM, the access
time was given as the combination of travel time to the nearest vertiport and an average waiting time of
ﬁve minutes (assuming a service integrated with public transportation and a 10–min frequency4). In–vehicle275
travel times (including vehicle take–oﬀ and landing) were based on a travel speed approximation of 150 km/h
and the direct distance between the origins and destinations. Boarding time was approximated as the sum of
both boarding and de–boarding times, assumed to be 5 minutes in total. On the other hand, taxi attributes
were based on traﬃc conditions and a waiting time of ﬁve minutes was assumed for taxi; service scenarios
are presented in Figure 1. Following this introduction, respondents were ﬁrst asked to rank the given factors280
they believed to be the most important for adopting and using UAM. These were “booking experience”,
“on-time performance (reliability)”, “operation characteristics (availability/frequency of service)”, “process
of boarding”, “safety”, “trip cost”, “trip purpose”, “trip duration”, “vehicle characteristics (comfort and
cleanliness)”. Then, respondents’ perceived usefulness of the service was assessed in a Likert-scale question
with options ranging from “not at all useful” to “extremely useful”.285
In this second part of the survey, latent constructs aﬀecting automation acceptance were investigated,
in the form of ﬁve–point Likert scale agreement statements (ranging from “strongly disagree” to “strongly
agree”), such as data concerns, trust and safety, cost perception, and the value of travel time savings.
Statements on data concerns were: “I am worried that my data goes to a third party”, “My fear of cyber–
security could prevent me from using UAM”, “I am concerned about the loss of jobs induced by automation”;290
on trust and safety through UAM’s operation characteristics: “Service reliability (on-time performance) is
a very important feature for trusting UAM”, “In order for me to feel safe, I would expect UAM’s vehicles
to be equipped with surveillance cameras”, “I should be able to talk to an operator on the ground at any
time”, “The operator should be able to override the system and remotely control the UAM’s vehicles in case
of emergency”, “The service provider’s reputation is very important to gain trust to use UAM”. Statements295
on cost perceptions included the following: “I would be willing to use the service, as long as its price is
in the same range as that of a taxi”, “I would ﬁrst think about cost when deciding whether or not to use
UAM”, “I think UAM’s service costs provided in the scenarios are reasonable”. Finally, statements on time
savings were as follows: “Travel time saving is a key factor in deciding whether or not to use UAM”, “5-min
travel time saving is important”, “10-min travel time saving is important”, “20-min travel time saving is300
Latent statements were followed by the stated preference question on adoption time horizon: “When
are you most likely going to use UAM?”. Alternatives included options ranking from the ﬁrst year of
4Both access and egress time were based on assumptions of a 10–min frequency of the service.
Figure 1: Survey scenarios
operation (Y1), the second or third years of operation (Y2–Y3), the fourth or ﬁfth years of operation (Y4–
Y5), starting the sixth year of its operation (Y6+), to “never” (non–adopters), and “unsure” (uncertain305
adopters). Respondents were also asked about the most probable trip purpose they would use UAM for,
including “daily commute”, “business travel”, and “leisure”.
The third part of the survey also aimed at revealing latent variables on the respondents’ social atti-
tudes, through similar Likert-scale constructs. The familiarity with various on–demand services (Airbnb,
DriveNow/Car2Go, Uber, BlablaCar: “I am not familiar with the service” to “I use it frequently”), the use310
of social media platforms (Facebook, WhatsApp, Instagram, Twitter: “I don’t use it” to “I use it several
times a day”), the comfort with online services (online booking, banking, shopping: “very uncomfortable”
to “very comfortable”), with ﬂying (including “Prefer not to answer”), the willingness to share a ride (such
as in a taxi or BlaBlaCar: “very unwilling” to “very willing”) with strangers, the enjoyment of driving a
car (“don’t enjoy it at all” to “very much enjoy it”), and environmental awareness were investigated. The315
latter included ﬁve–point Likert scale agreement statements: “I am concerned about global warming”, “I
do not change my behavior based on environmental concerns”, “I am willing to spend a bit more to buy a
product that is more environmentally friendly”. Respondents were also asked if they had been involved in a
car crash and its severity level (if yes, between “no injuries”, “minor injuries”, and “major injuries”). The
survey ended with a free comment ﬁeld where respondents could optionally express their further concerns320
First, a pilot survey was conducted to gain useful insights on potential biases in the survey design. The
insights gained from this pilot study (in the form of comments and by evaluating preliminary models) were
incorporated in the ﬁnal survey, which was then published online using Limesurvey Pro (limesurvey.org).
The survey was publicly available for two months starting from the 18th of July 2018 and disseminated325
through various channels, including Facebook, Instagram,and mailing lists (such as student lists at the
Technical University of Munich).
3.2. Descriptive analysis and model development
A preliminary descriptive analysis was performed to understand the sample distribution and the stated
adoption of diﬀerent demographics. Then, models were built using exploratory factor analysis and discrete330
choice modeling (namely multinomial logit models and ordered logit models).
The exploratory factor analysis (EFA) was applied to the second and third parts of the survey pertaining
to the respondents’ perceptions (mostly of UAM automation) and social behaviors, respectively. This
method was used to primarily identify latent constructs behind the variables and secondarily to reduce the
data dimensionality. The variables mainly used were those with the same scale, such as ﬁve–point Likert335
scale agreement statements. In this study, EFA was applied using the maximum likelihood estimation (MLE)
as a factor extraction method: factanal in the R statistical computing software (R Core Team,2019). The
suggested number of factors was obtained using the Kaiser–Guttman method (Yeomans & Golder,1982).
Also, after trying diﬀerent rotations, varimax orthogonal rotation was used (Kaiser,1958). Factor scores
were computed using factor.scores and the “component” method, a weighted sum of the factor loads.340
As the answer options of the survey are discrete categories of adoption time horizon, discrete choice
modeling was applied to identify and analyze the factors signiﬁcantly aﬀecting adoption (independent vari-
ables). Using pythonbiogeme (Bierlaire,2003), both multinomial and ordered logit models were speciﬁed
and estimated. As the outcomes for “Y6+” and “never” represented very small percentages of the sample
size (less than 5%), both categories were ultimately merged. Starting from multinomial logit models and345
based on the inputs from the factor analysis, models were developed in a stepwise fashion, ﬁrst backward
(from saturated models), where only variables of high signiﬁcance (95% or 90 % conﬁdence interval) were
kept, then forward (from empty models), where signiﬁcant variables were added one after the other, using
similar conﬁdence levels. Both generic and alternative speciﬁc models were developed, where independent
variables were speciﬁc to the diﬀerent outcomes. Hypothesis testing was used in case some variables had350
similar estimates in order to test if merging them into generic variables would improve the overall model.
Also, models were compared against each other by performing a log–likelihood test, and by assessing the
values of the statistical parameters AIC and BIC. After assessing the results of the multinomial logit mod-
els (MNLs), ordered logit models (OLMs) were proposed by initially specifying models with variables that
showed patterns across the outcomes (adoption time horizon), then testing them for signiﬁcance like in the355
MNLs. After testing the model performance, the late adopters (Y6) and non–adopters (never) were merged
in the proposed OLM. Despite the limitations of OLMs for the non–ordinal scale, these were used for testing
the possible order of the outcomes, mostly for “unsure” respondents.
3.3. Study limitations
The convenience sample resulting from the online dissemination of the survey is a limitation of the360
sampling approach, might inﬂuence the results, and may potentially lead to a lower representativeness of
the population. To overcome this, the authors conducted statistical tests to test the signiﬁcance of attitudes
across demographics; for instance gender, or language. These were then taken into consideration for the
model development. Moreover, the assumption that factors used for the factor analysis are uncorrelated is
subject to limitations; Fabrigar et al. (1999) advise instead the use of oblique correlations, and Mokhtarian365
et al. (2009) use it in their study to reveal correlations among factors.
Since this is a Stated Preference (SP) survey, some statements might be biased. Particularly, attitude
statements may induce some prompting due to their formulation, as they lead to respondents’ tendency to
agree, also known as the acquiescence bias. According to Baumgartner & Steenkamp (2001), this response
bias is a threat to the validity of the questionnaire results. To overcome this, these statements were controlled370
for their quality, as discussed in Section 4.
Finally, most factors in our study are perception variables, only related to respondents’ attitudes and
not to the service attributes, which might aﬀect the dataset variability.
4. Data collection and results
The survey generated 221 responses, with most respondents residing in Europe (181 respondents), par-375
ticularly in Germany (138 or about 60% of the sample size), and speciﬁcally the region of Munich (97
respondents). The rest were scattered outside Europe between the US, Latin America, and the Middle
East. Due to its high concentration and signiﬁcance, the group of Munich was considered as a signiﬁcant
subsample. It should be noted that the location only represents the place of residence, and not the place of
origin. The average completion time was 13.6 min with a standard deviation of 6.5 min. Overall, a good380
representation of gender, occupation, household size, and household incomes was observed, as well as an
overrepresentation of the age category 25–34 and of higher levels of education, possibly due to the online
method of survey distribution. Also, public transportation as a main commute mode was overrepresented
in the Munich subsample. In all cases, missing values (representing less than 5% of the sample size) were
recoded using mean or median values (Tabachnick et al.,2007). The summary statistics of the entire sam-385
ple and the major subsample (Munich) are presented (before further processing) in Table 1, benchmarked
against the latest Munich Census for reference (Statistische ¨
Amter des Bundes und der L¨ander,2014).
Among the 221 respondents, 36.65% stated that they would adopt UAM in the second or third year of
its implementation, followed by 22.17% claiming an adoption during its ﬁrst year, 14.03% during its fourth
and ﬁfth year, 2.71% starting its sixth year, and 3.17% stating that they would never adopt the service.390
Moreover, 21.27% of the respondents expressed uncertainty (“unsure”) on their adoption time horizon of
UAM. The analysis of the results highlighted the importance of safety for UAM adoption as the majority
of respondents (more than 50 %) ranked it as the most important factor in their intention to adopt UAM.
Also, a strong indication towards the importance of UAM costs (second factor), trip duration (third factor),
on–time reliability (fourth factor), and operation characteristics (ﬁfth factor) was given. The suggested395
ranking was obtained by extracting for each rank (e.g.: ﬁrst rank, second rank, etc.) the factor with the
highest percentage of respondents. For the second to ﬁfth ranks, the mentioned factors had a share of about
20 % of the respondents. On the other hand, vehicle characteristics, boarding process, booking experience,
and trip purpose were ranked as sixth to ninth factor by most of the respondents, gathering each also about
20 % of the respondents.400
Table 1: Summary of sample and subsample characteristics
Gender Female 43.0% 51.6% 48.6%
Male 56.1% 47.4% 51.4%
Prefer not to an-
0.9% 1.0% -
Agea0-17 0.5% - -
18-24 19.5% 23.7% 9.2%
25-34 45.7% 56.7% 21.7%
35-44 19.0% 16.5% 22.4%
45-54 9.5% 16.5% 22.2%
55-64 5.0% - 16.8%
65+ 0.9% - 7.7%
Main occupation Full–time employed 57.9% 46% 87.1%
Part-time employed 9.1% 8%
Student 28.1% 42% 2.9%
Unemployed 0.5% 1% 2.2%
Self-employed 2.3% 1% 7.8%
Retired 0.9% -
Prefer not to an-
1.4% 1% -
Education High School 8.6% 6% 34.1%
Apprenticeship 2.7% 1% 40.7%
Bachelor 26.7% 26% 22.7%
Master 47.5% 52%
Doctorate 13.1% 13% 2.5%
Prefer not to an-
1.4% 2% -
Household incomeb<500 AC 7.2% 13.4 %
500-1000 AC 8.6% 13.4%
1000-2000 AC 11.3% 10.3%
2000-3000 AC 14.0% 16.5%
3000-4000 AC 10.9% 6.2%
4000-5000 AC 10.9% 7.2%
5000-6000 AC 4.9% 3.1%
6000-7000 AC 5.4% 30.9%
>7000 AC 6.3% 21.7%
Prefer not to an-
Main commute cCar as a driver 33.9% 14.0% 31.0%
mode Car as a passenger 1.8% -
40.7% 63.0% 28.8%
Bike 15.4% 20.0% 14.7%
Walk 5.9% 1.0% 25.5%
Other 2.3% 2.0% -
aCensus values for age are based on diﬀerent age categories and are therefore best-ﬁt values for the survey classes.
bThere are no census values for income levels; however the average disposable household income for the year 2016 is
4220 AC (Euromonitor International,2017).
cThe given values are trip distribution percentages (MVG 2011).
A preliminary analysis of the attitudes of diﬀerent demographics showed the importance of these factors
for the adoption intention. This was mostly observable for gender, main occupation, education, survey
language, and previous crashes, as observed in Figure 2. Females were found to have a lower tendency of
being immediate adopters (Y1), and a higher of being “unsure” about their adoption time; as expressed by
the 12% of female respondents for Y1 compared to 30% for their male counterparts. For the main occupation,405
respondents working full–time expressed a higher immediate adoption intention (30%) compared with part–
time employees (0%), or students (14%), which can be due to higher income levels and/or higher value of
time. The survey language also seemed to impact stated adoption. Respondents who ﬁlled the survey in
English showed a higher interest in early adoption (31%) and less skepticism (12%) compared to those ﬁlling
the survey in German. Respondents with previous crashes and no injuries expressed a higher intention410
of immediate adoption (26%) compared to those who had injuries. Respondents with diﬀerent education
levels also stated diﬀerent preferences: those with lower levels of education (bachelor or lower) expressed
less uncertainty (19%) compared to the ones with higher levels of education (doctorate). On the other hand,
the latter group showed a higher interest in immediate adoption (26%) compared to the former (18%). Of
course some of these comparisons are limited to their statistical signiﬁcance, since the compared outcomes415
were not well represented for all categories, such as occupation, crash history, and education levels. On the
other hand, gender and survey language were rather well distributed.
Percentage adoption (%)
Stated time adoption Y1 Y2−Y3 Y4−Y5 Y6+ Never Unsure
Figure 2: UAM adoption by diﬀerent demographics
The exploration of respondents’ attitudes, based on the Likert scale statements, showed that gender and
survey language were the demographics where most diﬀerences were observed. As presented in Figure 3,
females seemed to give a higher level of importance to the loss of jobs due to automation as well as to safety;420
the latter is demonstrated by higher safety camera expectations and requirements for an operator on the
ground and to override the vehicle in case of emergency. On the other hand, male respondents seemed to
have more ﬂying comfort than their female counterparts. These ﬁndings are statistically signiﬁcant (error
bars in Figure 3drawn based on the standard error) and compatible with those from the literature; in
particular the study by Hohenberger et al. (2016), where women were found to have a lower intention to use425
autonomous vehicles, possibly due to the eﬀect of gender on anxiety, rather than on pleasure.
Diﬀerent attitudes were observed based on the survey language. Notably, the respondents ﬁlling the
survey in English expressed a higher trust in automation, enjoyment of automation, perceived usefulness of
ADAS, and of UAM, compared to those ﬁlling the survey in German. A similar, less remarkable, pattern was
observed for respondents residing in Munich as opposed to those who don’t. This ﬁnding parallels the results430
of the study on Autonomous Vehicles Readiness, where Germany was ranked 12th in consumer acceptance,
despite being third in technology and innovation (KPMG,2018): it can be assumed that language reﬂects
better the culture than the place of residence.
Loss of job concerns
Operator on ground
Mean value on a 5−point Likert scale
All significant to at least 97% confidence level
Gender Female Male
Fear of cyber−
0 1 2 3 4
Mean value on a 5−point Likert scale
All significant to at least 99.9% confidence level,
except for cyber−security with a 94% confidence
Survey language English German
Figure 3: Attitudes towards diﬀerent statements: (a) by gender; and (b) by language
Attitude statements from the second and third parts of the survey (see Section 3.1) were also analyzed
to assess the quality of the responses. After a careful examination of the distribution of the answers across435
the Likert–scale (“strongly disagree” to “strongly agree”), two statements were identiﬁed as prompted. This
was observed for the agreement statements on the necessity of an operator to override the vehicle in case
of emergency, and on the importance of the service provider’s reputation to gain trust to use UAM. For
both statements, the respondents’ tendency to agree, also known as the acquiescence bias, was noted, as
mentioned in the study limitations (see Section 3.3). Accordingly, both statements were removed from the440
subsequent analysis and model speciﬁcations.
Finally, a qualitative analysis of the comments left at the end of the survey showed additional concerns
of the respondents. The biggest concern was expressed by eight respondents on environmental impacts,
including concerns from noise and visual pollution (four respondents for each). For the latter, the respondents
expressed the disturbance they would feel from being watched from above (“Not in My Backyard” eﬀect,445
as described by one respondent). Two respondents expressed their concerns about privacy (concerning data
sharing) and safety (one of them stated the desire to wait for an incident–free operation of UAM). Moreover,
three respondents expressed their concern of economic impacts of UAM; one of them was concerned that the
service would target a niche market. Also, two respondents expressed their desire for a system integration of
UAM (with the existing and future transportation systems), one respondent was interested in the information450
sharing inside the vehicle, another for a high frequency service with on–demand availability, and ﬁnally one
more about the use case or purpose of UAM.
5. Model development and estimation
This section describes the model development, starting from exploratory factor analysis and continuing
with the developed discrete choice models.455
5.1. Exploratory factor analysis
The outcome of the exploratory factor analysis for respondents’ perceptions on automation is presented
in Table 2. Four factors were eventually extracted, explaining a cumulative variance of 52% of the total
variance. As observed in Table 2, each factor explains more than 10% of the total variance (with the exception
of one that explains 9%5), which is considered acceptable according to Costello & Osborne (2005). This is460
compatible with the work of several researchers who used factor analysis to understand users’ perceptions
of transportation systems (Tyrinopoulos & Antoniou,2008;Efthymiou et al.,2013), reducing the initial
indicators to fewer factors each explaining more than 10% of the total variance (with one exception at
most), with a cumulative total variance ranging roughly from an average of 46% to 50%. Loadings lower
than 0.4 are not shown to simplify the table.465
By looking at the latent meaning that these variables might have, the factors are interpreted as follows.
The value of time savings is a cluster of the three variables on the perception of saving 5, 10, and 20 min
of travel time. The enjoyment of automation6, trust of automation, and perceived usefulness of UAM are
interpreted as the aﬃnity to automation. The fear of cyber–security, the concern of data being shared to a
third party, and the concern of loss of jobs due to automation are grouped under data and ethical concerns.470
Finally, the need for an operator on the ground, and of in–vehicle safety cameras are interpreted as safety
Table 2: Factor analysis on respondents’ perceptions
Loadings Factor 1 Factor 2 Factor 3 Factor 4
Travel time savings 5min 0.78
Travel time savings 10min 0.98
Travel time savings 20min 0.59
Enjoy automation 0.80
Trust automation 0.75
UAM is useful 0.51
Fear of cyber–security 0.70
Fear that data goes to a third party 0.53
Loss of job concerns 0.49
Operator on the ground 0.67
In–vehicle safety cameras 0.50
Sum of square of loadings 2.01 1.66 1.14 0.96
Proportion variance 0.18 0.15 0.10 0.09
Cumulative variance 0.18 0.33 0.44 0.52
Factor interpretation Value of
For respondents’ social attitudes, four factors were extracted, explaining a cumulative variance of 55% of
the total variance, presented in Table 3. Variables pertaining to social attitudes were grouped under aﬃnity
to online services, environmental awareness, aﬃnity to social media, and aﬃnity to sharing, with all factors475
explaining at least 10% of the a variance and therefore also considered acceptable (Costello & Osborne,
5This factor was retained due to its high explanatory power and interpretability.
6Enjoyment of automated systems like Alexa or Siri
Table 3: Factor analysis on respondents’ social attitudes and behaviors
Loadings Factor 1 Factor 2 Factor 3 Factor 4
Online booking 0.81
Online banking 0.86
Online shopping 0.65
Concerned about global warming 0.99
Spend on environmental products 0.5
Willingness to share 0.5
Sum of square of loadings 1.93 1.28 1.24 1.06
Proportion variance 0.19 0.13 0.12 0.11
Cumulative variance 0.19 0.32 0.45 0.55
Factor interpretation Aﬃnity to
The factor analysis interpretation (Tables 1 and 2) is similar to the one on shoppers’ attitudes by
Mokhtarian et al. (2009), where a factor analysis was applied to 42 attitude agreement statements (ﬁve-
point Likert Scale), reducing them to 13: shopping enjoyment, time consciousness, trust, trendsetting,480
pro-environment, technology; these are comparable to the presented factors: aﬃnity to automation, time
savings, safety concerns, aﬃnity to online services and social media, environmental awareness, and aﬃnity
to automation and social media, respectively.
5.2. Multinomial logit model
Aiming at a better understanding with regards to the factors which aﬀect adoption, a multinomial logit485
model was estimated; the model speciﬁcation and estimation results are shown in Table 4. The coeﬃcient
estimates were in general reasonable in sign and magnitude and consistent with prior expectations; most of
them signiﬁcant at least to the 95% conﬁdence level with a t–value higher than 1.96, with a couple being
signiﬁcant to the 90% with a t–value higher than 1.65. For instance, estimates for the aﬃnity to automation
were highly signiﬁcant and positive mostly for early adoption (estimate of 1.15 for Y1), but also (to a lesser490
extent) for later years (estimate of 0.79 for Y2–Y5). For most of the respondents (Y1–Y5 and unsure),
highly signiﬁcant and positive impacts were also noted for the aﬃnity to social media and to WhatsApp
(estimate coeﬃcients of 0.50 and 0.53, respectively); the same impact was observed for full–time employment
(estimate of 0.63 for Y1–Y3). Cost factors also had a signiﬁcant positive impact on adoption. Scaling UAM
prices to taxis’ highly contributed to immediate adoption (estimate of 0.32 for Y1). On the other hand,495
higher data and ethical concerns (including fear of cyber–security, concerns of data being passed to third
parties, and loss of job concerns) had inﬂuential negative impacts on early adoption (-0.24 for Y1). Previous
crash experiences also had a negative impact on adoption in general, with lesser impacts on later adoption
(estimate of -2.42 for Y1, compared to -1.89 for Y2–Y5). Safety concerns (including the importance of having
an operator on the ground at any time and of in–vehicle safety cameras) also negatively aﬀected adoption500
(estimate of -0.34) with a high inﬂuence on immediate adoption (Y1) and later adoption (Y4–Y5). Female
respondents were in general less likely than their male counterparts to adopt UAM (estimate of -2.94 for all
years) and highly educated respondents (doctorate level or higher) less likely to adopt it in its early years
(estimate of -1.43 for Y2–Y3). Moreover, German as a survey language (compared to English) was highly
correlated with uncertainty (estimate of 1.42). Furthermore, higher income level respondents were less likely505
to be late adopters of UAM (estimate of -1.42 for Y4–Y5). The perception of travel time importance was
also decisive in UAM adoption for the second and third years (estimate of 0.28 for Y2–Y3) and the value
of time savings (including the perception of each of 5,10, and 20 minutes of time savings) for later adoption
(estimate of 0.33 for Y4–Y5). Finally, public transport commuters were more likely to adopt UAM during
its later years (estimate of 0.85 for Y4–Y5) and the alternative–speciﬁc constant for respondents stating a510
very late or non–adoption (Y6+, Never) was positive and highly signiﬁcant (estimate of 6.73).
Table 4: Multinomial logit model results (N=221)
Parameters estimate std. error t–stat
ASC Y6+ or never 6.73 2.20 3.06
Aﬃnity to automation Y1 1.15 0.32 3.59
Aﬃnity to automation Y2–Y5 and unsure 0.79 0.31 2.60
Cost as taxi Y1 0.32 0.18 1.83
Previous crash with injuries Y1 -2.42 0.97 -2.49
Previous crash with injuries Y2–Y5, unsure -1.89 0.86 -2.20
Starting language German unsure 1.42 0.34 4.13
Data concerns Y1 -0.24 0.12 -1.98
Doctorate level of education Y2–Y3 -1.43 0.53 -2.67
Female Y1–Y5 and unsure -2.94 1.14 -2.58
Full–time employment Y1-Y3 0.63 0.31 2.09
High income: 3000–7000 Euro Y4–Y5 -1.42 0.65 -2.20
PT commute Y4–Y5 0.85 0.40 2.11
Safety concerns Y1 and Y4–Y5 -0.34 0.10 -3.46
Aﬃnity to social media Y1–Y5 and unsure 0.50 0.29 1.71
TT important for UAM Y2–Y3 0.28 0.07 3.81
Value of time savings Y4–Y5 0.33 0.08 4.07
Aﬃnity to WhatsApp Y1–Y5 and unsure 0.53 0.26 2.07
5.3. Ordered logit model
Ordered logit models were also built with adoption time horizon as a dependent variable. Although
time frame can theoretically be ordered from the ﬁrst to the sixth or more years (even never), the category
“unsure” cannot be ranked in that scale. Therefore, based on previously observed patterns in the MNL515
model, OLMs were proposed . For instance, the aﬃnity to automation estimate from the MNL (Table 4)
was positive and strongly signiﬁcant (more than 95%) for both immediate adoption (Y1) and later adopters
(Y2 to Y5) and unsure respondents; however, the impact was higher for immediate adopters (Y1), expressed
by the higher coeﬃcient estimate (1.15 compared to 0.79). Similarly, previous crashes with injuries negatively
impacted adoption, expressed by strongly signiﬁcant (more than 95%) and negative coeﬃcient estimates;520
higher even for Y1 (-2.42 compared to -1.89 for the rest). These for instance guided the speciﬁcation of
an ordered logit model, where such attributes would be ranked, as suggested by the patterns in the MNL.
Accordingly, two cases were proposed. In the ﬁrst case, the ordered categories were as follows: Y1; Y2–Y3;
Y4–Y5; Unsure; Y6+/Never. The second case was built from the ﬁrst by merging the “unsure” category
with the one of late and non–adopters. The corresponding categories were therefore: Y1; Y2–Y3; Y4–Y5;525
Unsure, Y6+ and Never.
Table 5represents the ﬁnal OLM model for both cases. The highly signiﬁcant cut values for case 1
indicate that adoption is indeed ordered and people who are unsure display a behavior that is ranked
between late (Y4–Y5) and extremely late or non–adopters (Y6+ or Never). This was rather expected from
the patterns observed in MNL models. The signiﬁcant parameters are the same for both cases: aﬃnity to530
automation, full–time employment, and cost as taxi (meaning the willingness to use the service for prices
in the range of taxis) are associated with an early adoption. On the other hand, data and ethical concerns
and starting language German are strongly correlated with a later adoption.
Table 5: Ordered logit models results (N=221)
Case 1 Case 2
Parameters estimate std. error t–stat estimate std. error t–stat
Aﬃnity to automation -0.26 0.07 -3.54 -0.23 0.07 -3.11
Cost as taxi -0.37 0.13 -2.82 -0.38 0.14 -2.70
Starting language German 0.61 0.30 2.05 0.70 0.30 2.32
Data and ethical concerns 0.27 0.08 3.32 0.26 0.08 3.19
Full–time employment -0.96 0.26 -3.68 -1.03 0.27 -3.84
Y1—Y2–Y3 -4.22 0.85 -4.97
Y2–Y3—Y4–Y5 -2.16 0.21 9.96
Y4–Y5—Unsure -1.36 0.14 5.93
Unsure—Y6+/Never 0.87 0.31 7.10
Y1—Y2–Y3 -4.02 0.88 -4.57
Y2–Y3—Y4–Y5 -1.96 0.21 9.92
Y4–Y5—Unsure/Y6+/Never -1.15 0.14 5.93
AIC 585.10 526.04
BIC 615.69 553.23
Compared to Case 1, Case 2 presents similar values for the relevant parameters with close estimates
in terms of value and signiﬁcance. This model also presents signiﬁcant cutoﬀ values between the ordered535
categories and indicates that uncertain respondents can be merged with very late (Y6+) and non–adopters.
In other terms, respondents showing some skepticism regarding their adoption time horizon of UAM are more
likely to use it at a later stage, or not to use it at all. In both cases, uncertain adopters are associated with
rather late adoption; the only diﬀerence is that one case considers uncertainty as part of non–adoption and
the other as one degree less. However, due to the diﬀerence in the nature of the data between these models540
(since the ﬁrst case provides ﬁve outcomes for the dependent variable compared to four in the second), a
likelihood ratio test is not possible. Therefore as both models have more or less the same meaning, deciding
on the number of ordered categories depends on personal judgment and preference: parsimony vs. richness.
6.1. Summary ﬁndings545
Despite high stated adoption rates for UAM (about 37% for Y2–Y3 and 22 % for Y1), a non-negligible
number of respondents expressed their uncertainty about their intention to use the service, as shown by the
percentage of “unsure” respondents (more than 22 %). The analysis of the responses showed that safety
was perceived as the factor of highest importance in using UAM (ranked by the majority of respondents
as the ﬁrst factor in UAM adoption). A high indication for the importance of trip cost, trip duration,550
service reliability and operation characteristics was also observed, as they were mostly ranked as second,
third, fourth, and ﬁfth most important factors for UAM adoption. The survey statistics indicated a high
impact of socio–demographic factors on adoption and attitudes. Females expressed a much lower interest in
immediate UAM adoption, expressing overall lower trust and perceived usefulness of automation, and greater
security and safety concerns and expectations, expressing a higher desire for operators on the ground, and555
of in–vehicle safety cameras, as reﬂected in the mean values of the attitude Likert–scale questions (partially
presented in Figure 3). On the other hand, both fully-employed and higher income respondents expressed
a greater interest in early adoption, as discussed in the survey statistics (Section 4.1). Cultural impact was
observed through the survey language, which seemed to also have an impact on automation; respondents
ﬁlling the survey in German expressed a lower interest in early adoption and a higher skepticism observed560
through a higher degree of uncertainty. This is interesting when compared to the impact of the place of
residence (Munich or Germany), which was less inﬂuential than the survey language, as only the latter was
a signiﬁcant variable in the adoption models (MNL and OLM).
The discussed ﬁndings from the statistics were conﬁrmed by the speciﬁed models. For instance, MNLs
demonstrated the high impact of gender (female), employment (full–time), language (German language),565
income levels on adoption. The MNLs also showed that public transportation users were more likely to be
late UAM adopters, as expressed by the positive and signiﬁcant coeﬃcient estimate for PT commuters (Table
4). Moreover, respondents with previous crash experiences with injuries were less likely to be immediate
adopters, as revealed by the model ﬁnding estimates (Table 4). Safety concerns were also found to play
an inhibiting role in early and late adoption, expressed by the negative and highly signiﬁcant (more than570
99.9 %) coeﬃcient estimate in the MNL. These included the need for in–vehicle surveillance cameras, and
for an operator on the ground at any time (resulting from the factor analysis in Table 2). Lack of safety
may hinder UAM, and lead to late UAM adoption. Moreover, the aﬃnity to automation was found crucial
for UAM adoption (as demonstrated by the positive coeﬃcients for the MNL, higher still for Y1 than for
Y4-Y5, and negative for the OLM, proving as well the higher eﬀect on immediate adoption of the aﬃnity575
to automation) and included the enjoyment and trust of automation, and the perceived usefulness of UAM.
Furthermore, the value of time savings and perceived costs were highly inﬂuential for late UAM adoption.
As some of these ﬁndings might be counter–intuitive, their occurrence might be due to respondents’ prior
expectations and judgments of service properties. For instance, respondents may be skeptical about actual
time savings and therefore as they value this factor, they might decide to adopt UAM later on, waiting for580
the service to improve its performance. This was also observed for respondents who are highly sensitive
to costs as they might believe that costs would decrease a few years after the service implementation to
be in the range of taxis. Moreover, higher safety concerns were negatively associated with late adoption,
possibly due to the users’ beliefs that safety would only need a few years to be demonstrated; waiting for too
long would therefore not be needed. Also, data and ethical concerns were found to be hindering immediate585
adoption; this could refer to the respondents who are rather worried about their data being shared to third
parties, and who tend to be uncertain about the future as they are concerned from cyber–security threats
and/or the loss of jobs due to automation. Finally, the impact of social media and WhatsApp was found to
be positive on adoption.
OLMs conﬁrmed some of these factors found in the MNLs, notably the cost range importance (comparable590
to the taxi range for early adoption), the cultural impact (correlation between German language and later
adoption), the data and ethical concerns (associated with later adoption and including the fear of cyber–
security, data sharing, and the loss of job concerns), the aﬃnity to automation, and the employment impact.
Finally, the ordered models gave meaningful insights regarding “unsure” respondents, in which these were
found to have similar or close (one degree less) behavioral intentions to late adopters.595
When comparing the results of this study with the reports commissioned by NASA and presented in
Section 2.1, parallel ﬁndings were observed. The study by Crown Consulting (NASA,2019) mentioned
that nearly half of the consumers were potentially comfortable with delivery and UAM use cases, which
is quite comparable with our results (more than 50% stated time adoption within the ﬁrst three years of
implementation). Moreover, common factors were observed such as safety, privacy, loss of jobs concerns,600
environmental concerns, and noise and visual pollution that were extracted from our qualitative analysis.
Common results were also shared with the study by Booz Allen Hamilton (NASA,2018), particularly for
user concerns like safety, privacy and noise, piloted aircraft and ﬂight attendant, cybersecurity, cost and
convenience, and gender impact with men being more comfortable and willing to use UAM; in our study
however, the system was assumed to be fully automated, and therefore a ﬂight attendant was not mentioned.605
Instead, the importance of having an operator (available on the ground and/or remotely control the vehicle
in case of emergency) was investigated.
6.2. Technology Acceptance Model for Disruptive Transport Technologies
The model ﬁndings pertaining to acceptance were incorporated to extend the Technology Acceptance
Model for the application of urban air mobility, and more generally disruptive transport technologies (Fig-610
ure 4). Compared to the initial TAM by Davis et al. (1989), the model retains only one of the two main
constructs: the perceived usefulness (PU), which directly impacts the behavioral intention (BI). The per-
ceived ease of use (PEU; in the case of urban air mobility: service booking or boarding) was not found
inﬂuential and was therefore removed from the model. More generally, for fully-automated systems, the
user is not expected to do any eﬀort. Actual system use is also omitted, as there is no way to measure it615
for non–existing systems. The attitude towards using the system is removed; instead the following factors
(including PU) are considered as attitudes directly impacting the behavioral intention: perceived usefulness,
social behavior (social attitudes related to social media use, environmental concerns, etc.), value of time
(the value given to time savings), perceived costs (cost perceptions: cost range), data and ethical concerns
(fear of data sharing to a third party, fear of cyber–security, and the loss of job concerns). These factors620
result from the estimated multinomial logit models (MNLs); data and ethical concerns and cost perceptions
also result from the ordered logit models (OLMs).
The construct of trust is added as a main construct in this model, as suggested by the Automation
Acceptance Model (AAM) by Ghazizadeh et al. (2012). Additionally, the proposed model introduces factors
directly and positively impacting trust. These are the perceived reliability of automation, the perceived625
vehicle’s safety, the perceived locus of control, and the previous experience with automation. The vehicle’s
safety and perceived locus of control are supported by the MNLs and refer to the safety from in–vehicle
safety cameras and the locus of control conﬁrmed by the importance given to the human factor (operator
on the ground at any time). Moreover, both service reliability and automation experience were found from
analyzing respondents’ attitudes. Finally, external variables, similarly to the AAM directly impact trust,630
but not PU. These are given as socio–demographics and aﬃnity to automation. The former includes gender,
education, occupation, income, cultural impact (demonstrated by the survey language). In the proposed
model, these external variables directly impact behavioral intention as well.
Trust/ Value of
Value of Time
Affinity to Automation
Data and Ethical
Locus of Control
Figure 4: Technology Acceptance Model for Disruptive Transport Technologies, adapted from the original TAM by Davis et al.
(1989) and the Automation Acceptance Model (AAM) by Ghazizadeh et al. (2012)
6.3. Insights for policymakers and industrial stakeholders
The ﬁndings of this research provide insights for both responsible policymaking and industrial stake-635
holder engagement, to ensure a smooth urban air mobility implementation and integration with the existing
transport systems. Recommendations on such areas can be summarized in the following points.
6.3.1. Insights for policymakers
As highlighted in the proposed Technology Acceptance Model, trust or the value of safety is a crucial640
component in users’ adoption of urban air mobility and is determined by several components. Among
these, the perception of automation reliability, vehicle’s safety and locus of control are areas where
stringent regulations could improve users’ trust and reduce skepticism resulting from missing or erro-
neous information. Accordingly, safety standards enforced by the responsible authorities could ensure
users’ trust, alleviating negative perceptions on automation reliability and service performance. En-645
forcing regulations acting upon in-vehicle safety, such as the presence of in-vehicle surveillance cameras,
might help increasing trust and feeling of safety. Finally, regulations focusing on the human factor in
the operation of urban air mobility could take into consideration the importance of the availability (at
any time) of an operator on the ground, to increase the users’ perceived locus of control, and thereby
trust and safety towards the service.650
The proposed TAM also highlights the importance of data and ethical concerns for UAM adoption,
and their negative impacts on safety perception. These include concerns from information sharing
to third parties, from cyber-security, and from the loss of jobs due to automation, as shown by the
exploratory factor analysis results. Responsible policymaking in the area of data sharing and protection655
would probably ensure a higher sense of privacy for the users. For ground Autonomous Vehicles,
Fagnant & Kockelman (2015) recommend the creation of nationally recognized licensing framework
for autonomous vehicles (ground), determining appropriate standards for liability, security, and data
privacy. Such frameworks could be appropriately developed and transferred to urban air mobility.
Not directly stemming from the models, environmental concerns expressed by respondents demonstrate
the need for policy–making in the area of noise and visual impact. Regulating the allowed noise levels
and ﬂying altitudes of the vehicles could address community acceptance of both users, and non-users
of urban air mobility. Moreover, an inclusive and integrated system, as deﬁned by our study should
ensure a proper integration of UAM services with existing transportation systems.665
6.3.2. Insights for industrial stakeholders
The proposed TAM revealed the importance of service attributes, i.e. cost and time, for the adoption of
urban air mobility. If the system aims to attract a wider public, user requirements must be prioritized,
notably pricing schemes must be deﬁned, and regulated by the authorities, so that the service would670
not only be a niche market. These for instance need to be within a certain reasonable range, i.e. taxi
prices, allowing competition with existing ground vehicles. Similarly, the time from access to egress
(to and from the vertiports) must be optimized: a smoother process must enable an eﬃcient process
for the service to be accepted; again, if the service aims to be integrated with the existing systems, this
can only be done through a proper stakeholder engagement with the responsible authorities. Finally,675
a higher transparency on both time and costs of the service would ensure a higher awareness on the
service attributes, which is crucial for users’ perceptions towards it.
For an inclusive service, industrial stakeholders should consider the importance of socio-demographics
in the perception of the service. User categories with diﬀerent gender or cultural background could be680
targeted according to their speciﬁc needs. Incentives might as well help some respondents to overcome
their fears of automation.
In this paper, users’ acceptance and adoption of urban air mobility was assessed by analyzing an online
stated preference study that gathered 221 respondents from all over the world, with a subsample of 97 Munich685
residents. Using mostly factors from the literature on the acceptance of both technology and automation
(notably autonomous vehicles), and by projecting them to UAM, the survey aimed at uncovering many
hypotheses on this new mobility service and thereby also validated an extended technology acceptance
model for an urban air mobility context. The analysis of the survey data highlighted the importance
of socio–demographic parameters and their attitudes in adoption. Using the exploratory factor analysis,690
many variables were clustered into groups having a high explanatory power and thus reducing the dataset
dimensionality. These included the aﬃnity to automation, safety concerns, data and ethical concerns, the
value of time savings, in addition to social attitudes such as environmental awareness, the aﬃnity to social
media, online services, and sharing. Using the inputs from the factor analysis, signiﬁcant multinomial and
ordered logit models were developed with adoption time horizon as a dependent variable, which can be a695
considered as a theoretical contribution to the use case of ordered logit models with a non–conventional
ordinal dependent time variable (including non-ordinal parameters such as “unsure” and “never”). Mostly,
the inﬂuence of socio–demographics, cultural impact, aﬃnity to automation (including the enjoyment and
trust of automation) were shown. Trust and safety were found to be key components for UAM adoption; in
particular, the presence of in–vehicle cameras and operators, as well as performance expectancy in terms of700
service reliability and on–time performance were noted. Data and ethical concerns, the value of time savings
and costs, social attitudes including a high aﬃnity to social media were also found as highly inﬂuential for
UAM adoption. Finally, public transportation as a commute mode was found to be rather related to late
Overall, the study extracted relevant and highly signiﬁcant factors, extending accordingly the TAM705
model for UAM use, and contributing in ﬁlling the gap in research on this new service. The ﬁndings of this
work have strong insights for policymakers. Regulatory bodies should have departments targeting UAM
regulations and work closely with other countries to have some uniﬁed standards; for instance, a uniﬁed
European framework on suggested safety standards and allowed noise levels.
Most factors in the developed models are related to the perception and attitudes of respondents, and710
not to the service attributes; newer models could therefore be developed with alternative-speciﬁc variables.
Scenarios could for instance present travel times or costs for diﬀerent time frames. Future work could
therefore focus on uncertainties pertaining to the service attributes and business models. Another research
motivation could be to build a nested ordered logit model with one or more ordered nests; one of the nests
could comprise uncertain or non–adopters and the other the certain adopters including the ordered (time)715
adoption categories. Other studies could test the developed UAM TAM by applying a conﬁrmatory factor
analysis (CFA) and developing a hybrid latent class model, in which both analysis methods (factor analysis
and choice models) would be combined.
This work has been partly funded by the Bavarian Ministry of Economic Aﬀairs, Regional Development and720
Energy through the OBUAM project.
Airbus (2017). Rethinking urban air mobility. URL: https://www.airbus.com/newsroom/stories/
Airbus (2018). Urban air mobility - the sky is yours. URL: https://www.airbus.com/innovation/725
Urban-air- mobility-the- sky-is-yours.html.
Antcliﬀ, K. R., Moore, M. D., & Goodrich, K. H. (2016). Silicon valley as an early adopter for on-demand civil VTOL
operations. In 16th AIAA Aviation Technology, Integration, and Operations Conference (p. 3466).
Bansal, P., Kockelman, K. M., & Singh, A. (2016). Assessing public opinions of and interest in new vehicle technologies: An
Austin perspective. Transportation Research Part C: Emerging Technologies,67 , 1–14.730
Baptista, P., Melo, S., & Rolim, C. (2014). Energy, environmental and mobility impacts of car-sharing systems. empirical
results from lisbon, portugal. Procedia-Social and Behavioral Sciences,111 , 28–37.
Baumgartner, H., & Steenkamp, J.-B. E. (2001). Response styles in marketing research: A cross-national investigation. Journal
of Marketing Research,38 , 143–156. URL: https://doi.org/10.1509/jmkr.188.8.131.5240. doi:10.1509/jmkr.38.2.143.
Bierlaire, M. (2003). Biogeme: a free package for the estimation of discrete choice models. In Swiss Transport Research
Bimbraw, K. (2015). Autonomous cars: Past, present and future a review of the developments in the last century, the present
scenario and the expected future of autonomous vehicle technology. In Informatics in Control, Automation and Robotics
(ICINCO), 2015 12th International Conference on (pp. 191–198). IEEE volume 1.740
Bjørner, T. (2015). A priori user acceptance and the perceived driving pleasure in semi-autonomous and autonomous vehicles.
In European Transport Conference 2015.
Boucher, P. (2016). ‘You wouldn’t have your granny using them’: Drawing boundaries between acceptable and unac-
ceptable applications of civil drones. Science and Engineering Ethics,22 , 1391–1418. URL: https://doi.org/10.1007/
Chamata, J. (2017). Factors delaying the adoption of civil drones: A primitive framework. The International Technology
Management Review,6, 125–132.
Choi, J. K., & Ji, Y. G. (2015). Investigating the importance of trust on adopting an autonomous vehicle. International
Journal of Human-Computer Interaction,31 , 692–702.
Clements, L. M., & Kockelman, K. M. (2017). Economic eﬀects of automated vehicles. Transportation Research Record:750
Journal of the Transportation Research Board, (pp. 106–114).
Clothier, R. A., Greer, D. A., Greer, D. G., & Mehta, A. M. (2015). Risk perception and the public acceptance of drones. Risk
Analysis,35 , 1167–1183.
Cohen, M. M. (1996). The Vertiport as an Urban Design Problem. Technical Report SAE Technical Paper.
Connected Automated Driving EU (2017). New study on user acceptance of driverless cars released. URL: https:755
//connectedautomateddriving.eu/mediaroom/nes-study- user-acceptance- driverless-cars-released/.
Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory factor analysis: Four recommendations for getting the
most from your analysis. Practical Assessment, Research & Evaluation,10 , 1–9.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: a comparison of two theoretical
models. Management Science,35 , 982–1003.760
Deloitte Analytics Institute (2017). Autonomous driving in Germany how to convince customers. URL: https://www2.
deloitte.com/content/dam/Deloitte/de/Documents/consumer-industrial- products/Autonomous-driving- in-Germany_
Dolnicar, S. (2013). Asking good survey questions. Journal of Travel Research ,52 , 551–574.
Efthymiou, D., Antoniou, C., & Waddell, P. (2013). Factors aﬀecting the adoption of vehicle sharing systems by young drivers.765
Transport Policy,29 , 64–73.
Euromonitor International (2017). Munich city review. URL: http://www.euromonitor.com/munich-city-review/report.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of exploratory factor analysis
in psychological research. Psychological methods,4, 272.
Fadhil, D. N. (2018). A GIS-Based Analysis for Selecting Ground Infrastructure Locations for Urban Air Mobility. Master’s770
Thesis Technical University of Munich.
Fagnant, D. J., & Kockelman, K. (2015). Preparing a nation for autonomous vehicles: opportunities, barriers and policy
recommendations. Transportation Research Part A: Policy and Practice,77 , 167–181.
Fagnant, D. J., & Kockelman, K. M. (2014). The travel and environmental implications of shared autonomous vehicles, using
agent-based model scenarios. Transportation Research Part C: Emerging Technologies,40 , 1–13.775
Fu, M., Rothfeld, R., & Antoniou, C. (2019). Exploring preferences for transportation modes in an urban air mobility
environment: a Munich case study. Transportation Research Record: Journal of the Transportation Research Board,In
Gaggi, S. (2017). Driverless cars and user acceptance. URL: https://mobilitybehaviour.eu/2017/07/26/
driverless-cars- and-user- acceptance/.780
Garrow, L. A., German, B., Mokhtarian, P., Daskilewicz, M., Douthat, T. H., & Binder, R. (2018). If you ﬂy it, will commuters
come? A survey to model demand for e–VTOL urban air trips. In 2018 Aviation Technology, Integration, and Operations
Conference (p. 2882).
Ghazizadeh, M., Lee, J. D., & Boyle, L. N. (2012). Extending the technology acceptance model to assess automation. Cognition,
Technology & Work,14 , 39–49.785
Hohenberger, C., Sp¨orrle, M., & Welpe, I. M. (2016). How and why do men and women diﬀer in their willingness to use
automated cars? the inﬂuence of emotions across diﬀerent age groups. Transportation Research Part A: Policy and Practice,
94 , 374–385.
Jenkins, S., Linden, E., & Wittmer, A. (2018). Autonomous Mobility - An Application of the Technology Acceptance Model.
URL: https://cfac.unisg.ch/en/research-publications- consulting/cfac-working- paper-series.790
Kaan, J. (2017). User acceptance of autonomous vehicles: Factors & implications, . URL: https://repository.tudelft.nl.
Kaiser, H. F. (1958). The varimax criterion for analytic rotation in factor analysis. Psychometrika ,23 , 187–200.
KPMG (2018). Autonomous vehicles readiness index. URL: https://assets.kpmg.com/content/dam/kpmg/tw/pdf/2018/03/
KPMG-Autonomous- Vehicle-Readiness- Index.pdf.
Kreimeier, M., & Stumpf, E. (2017). Market volume estimation of thin-haul on-demand air mobility services in Germany. In795
17th AIAA Aviation Technology, Integration, and Operations Conference (p. 3282).
Krueger, R., Rashidi, T. H., & Rose, J. M. (2016). Preferences for shared autonomous vehicles. Transportation Research Part C:
Emerging Technologies ,69 , 343 – 355. URL: http://www.sciencedirect.com/science/article/pii/S0968090X16300870.
Kyriakidis, M., Happee, R., & de Winter, J. C. (2015). Public opinion on automated driving: Results of an international
questionnaire among 5000 respondents. Transportation Research Part F: Traﬃc Psychology and Behaviour,32 , 127–140.800
Lidynia, C., Philipsen, R., & Zieﬂe, M. (2017). Droning on about dronesacceptance of and perceived barriers to drones in civil
usage contexts. In Advances in Human Factors in Robots and Unmanned Systems (pp. 317–329). Springer.
Likert, R. (1932). A technique for the measurement of attitudes. Archives of psychology ,20 , 5–55.
MacSween, S. (2003). A public opinion survey-unmanned aerial vehicles for cargo, commercial, and passenger transportation.
In 2nd AIAA” Unmanned Unlimited” Conf. and Workshop & Exhibit (p. 6519).805
McFadden, D. (2001). Economic choices. American economic review,91 , 351–378.
Merat, N., Madigan, R., & Nordhoﬀ, S. (2016). Human factors, user requirements, and user acceptance of ride-sharing
in automated vehicles. In International Transport Forum Roundtable on Cooperative Mobility Systems and Automated
Driving (pp. 6–7).
Mokhtarian, P. L., Ory, D. T., & Cao, X. (2009). Shopping-related attitudes: a factor and cluster analysis of Northern California810
shoppers. Environment and Planning B: Planning and Design,36 , 204–228.
NASA (2018). Urban Air Mobility (UAM) Market Study. Technical Report. URL: https://ntrs.nasa.gov/archive/nasa/
NASA (2019). Urban Air Mobility (UAM) Market Study. URL: https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/
Nees, M. A. (2016). Acceptance of self-driving cars: an examination of idealized versus realistic portrayals with a self-driving
car acceptance scale. In Proceedings of the Human Factors and Ergonomics Society Annual Meeting (pp. 1449–1453). SAGE
Publications Sage CA: Los Angeles, CA volume 60.
Payre, W., Cestac, J., & Delhomme, P. (2014). Intention to use a fully automated car: Attitudes and a priori acceptability.
Transportation Research Part F: Traﬃc Psychology and Behaviour ,27 , 252–263.820
Peeta, S., Paz, A., & DeLaurentis, D. (2008). Stated preference analysis of a new very light jet based on-demand air service.
Transportation Research Part A: Policy and Practice,42 , 629 – 645. URL: http://www.sciencedirect.com/science/
Piao, J., McDonald, M., Hounsell, N., Graindorge, M., Graindorge, T., & Malhene, N. (2016). Public views towards imple-
mentation of automated vehicles in urban areas. Transportation Research Procedia,14 , 2168–2177.825
Porsche Consulting (2018). The future of vertical mobility: Sizing the market for passenger, inspection, and good services until
R Core Team (2019). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing
Vienna, Austria. URL: http://www.R-project.org/.
Riad, M.-I., Jaradat, m.-i., Fagih, K., & S Faqih, M. (2014). Investigating the moderating eﬀects of gender and self-eﬃcacy830
in the context of mobile payment adoption: A developing country perspective. International Journal of Business and
R¨odel, C., Stadler, S., Meschtscherjakov, A., & Tscheligi, M. (2014). Towards autonomous cars: the eﬀect of autonomy levels
on acceptance and user experience. In Proceedings of the 6th International Conference on Automotive User Interfaces and
Interactive Vehicular Applications (pp. 1–8). ACM.835
Rothfeld, R. L., Balac, M., Ploetner, K. O., & Antoniou, C. (2018). Initial Analysis of Urban Air Mobility’s Transport
Performance in Sioux Falls. In AIAA AVIATION, Modeling and Simulation for Unmanned and Personal Aerial Vehicle
Rychel, A. (2016). Cultural divides: How acceptance of driverless cars varies globally. URL: https://www.2025ad.com/latest/
Schoettle, B., & Sivak, M. (2014). A survey of public opinion about autonomous and self-driving vehicles in the US, the UK,
Shamiyeh, M., Bijewitz, J., & Hornung, M. (2017). A review of recent personal air vehicle concepts. In Aerospace Europe 6th
Amter des Bundes und der L¨ander (2014). Ergebnisse des zensus 2011. URL: https://www.zensus2011.de/DE/845
Steele, C. M. (1995). Stereotype threat and the intellectual test performance. Journal of Personality and Social Psychology,
69 , 1.
Straubinger, A., & Verhoef, E. T. (2018). (Working Paper) Options for a Welfare Analysis of Urban Air Mobility. Hong Kong.
Sun, H., & Zhang, P. (2006). The role of moderating factors in user technology acceptance. International Journal of Human-850
Computer Studies,64 , 53–78.
Tabachnick, B. G., Fidell, L. S., & Ullman, J. B. (2007). Using multivariate statistics volume 5. Pearson Boston, MA.
Thipphavong, D. P., Apaza, R., Barmore, B., Battiste, V., Burian, B., Dao, Q., Feary, M., Go, S., Goodrich, K. H., Homola, J.
et al. (2018). Urban air mobility airspace integration concepts and considerations. In 2018 Aviation Technology, Integration,
and Operations Conference (p. 3676).855
Tyrinopoulos, Y., & Antoniou, C. (2008). Public transit user satisfaction: Variability and policy implications. Transport
Policy,15 , 260–272.
Uber Elevate (2016). Fast-forwarding to a future of on-demand urban air transportation. URL: https://www.uber.com/
Uber Elevate (2018). Fast-forwarding to the future of on-demand, urban air transportation. URL: https://www.uber.com/860
Vascik, P. D. (2017). Systems-level analysis of On Demand Mobility for aviation. Ph.D. thesis Massachusetts Institute of
Vascik, P. D., & Hansman, R. J. (2018). Scaling constraints for urban air mobility operations: Air traﬃc control, ground
infrastructure, and noise. In 2018 Aviation Technology, Integration, and Operations Conference (p. 3849).865
Venkatesh, V., & Bala, H. (2008). Technology acceptance model 3 and a research agenda on interventions. Decision sciences,
39 , 273–315.
Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal ﬁeld
studies. Management Science,46 , 186–204.
Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a870
uniﬁed view. MIS quarterly, (pp. 425–478).
Wang, Y., Xia, H., Yao, Y., & Huang, Y. (2016). Flying eyes and hidden controllers: A qualitative study of people’s privacy
perceptions of civilian drones in the US. Proceedings on Privacy Enhancing Technologies,2016 , 172–190.
Yeomans, K. A., & Golder, P. A. (1982). The guttman-kaiser criterion as a predictor of the number of common factors. The
Statistician, (pp. 221–229).875
Zhang, T., Tao, D., Qu, X., Zhang, X., Lin, R., & Zhang, W. (2019). The roles of initial trust and perceived risk in publics
acceptance of automated vehicles. Transportation Research Part C: Emerging Technologies,98 , 207–220.