ArticlePDF Available

Abstract and Figures

Autonomous driving is being discussed as a promising solution for transportation related issues and might bring some improvement for users of the system. For instance, especially high mileage commuters might compensate for some of their time spent travelling since they will be able to undertake other activities while going to work. At the same time, there are still many uncertainties and few empirical data on the impact of autonomous driving on mode choices. This study addresses the impact of autonomous driving on value of travel time savings (VTTS) and mode choices for commuting trips using stated choice experiments. Two use cases were addressed – a privately owned and a shared autonomous vehicle – compared to other modes of transportation. The collected data were analyzed by performing a mixed logit model. The results show that mode-related factors such as time elements, especially in-vehicle time and cost, play a crucial role for mode choices that include autonomous vehicles. The study provides empirical evidence that autonomous driving may lead to a reduction in the VTTS for commuting trips. We found that driving autonomously in a privately owned vehicle might reduce the VTTS compared to driving manually and is perceived similarly to in-vehicle time in public transportation. Also, riding in a shared autonomous vehicle is perceived less negatively than driving manually. The study provides important insights on VTTS by autonomous driving for commuting trips and can be a base for future research to build upon.
Content may be subject to copyright.
1
Working Paper
HOW AUTONOMOUS DRIVING MAY AFFECT THE VALUE OF TRAVEL TIME SAVINGS
FOR COMMUTING
Felix Steck, Corresponding Author
German Aerospace Center, Institute of Transport Research
Rutherfordstrasse 2, 12489 Berlin, Germany
Email: Felix.Steck@dlr.de
Viktoriya Kolarova
German Aerospace Center, Institute of Transport Research
Rutherfordstrasse 2, 12489 Berlin, Germany
Email: Viktoriya.Kolarova@dlr.de
Francisco Bahamonde-Birke
German Aerospace Center, Institute of Transport Research
Rutherfordstrasse 2, 12489 Berlin, Germany
and
Technische Universität Berlin
Email: Francisco.BahamondeBirke@dlr.de; bahamondebirke@gmail.com
Stefan Trommer
German Aerospace Center, Institute of Transport Research
Rutherfordstrasse 2, 12489 Berlin, Germany
Email: Stefan.Trommer@dlr.de
Barbara Lenz
German Aerospace Center, Institute of Transport Research
Rutherfordstrasse 2, 12489 Berlin, Germany
and
Humboldt-Universität zu Berlin
Email: Barbara.Lenz@dlr.de
Revised version submitted to the Transportation Research Board (TRB) 97th Annual Meeting
Important note: This is a preliminary version of the paper. A reviewed and revised version will be
published in Spring 2018 in the Journal "Transport Research Record: Journal of the Transportation
Research Board".
Working Paper
2
ABSTRACT
Autonomous driving is being discussed as a promising solution for transportation-related
issues and might bring some improvement for users of the system. For instance, especially
high mileage commuters might compensate for some of their time spent travelling since they
will be able to undertake other activities while going to work. At the same time, there are still
many uncertainties and few empirical data on the impact of autonomous driving on mode
choices.
This study addresses the impact of autonomous driving on value of travel time savings (VTTS)
and mode choices for commuting trips using stated choice experiments. Two use cases were
addressed a privately owned and a shared autonomous vehicle compared to other modes
of transportation. The collected data were analyzed by performing a mixed logit model.
The results show that mode-related factors such as time elements, especially in-vehicle time
and cost, play a crucial role for mode choices that include autonomous vehicles. The study
provides empirical evidence that autonomous driving may lead to a reduction in the VTTS for
commuting trips. We found that driving autonomously in a privately owned vehicle might
reduce the VTTS by 31% compared to driving manually and is perceived similarly to in-vehicle
time in public transportation. Also, riding in a shared autonomous vehicle is perceived 10%
less negatively than driving manually. The study provides important insights on VTTS by
autonomous driving for commuting trips and can be a base for future research to build upon.
Keywords: autonomous driving, value of travel time savings, commuting trips, discrete
choice experiment, mixed logit
Working Paper
3
INTRODUCTION
Digitalization trends and rapid technology development have increased automation in all
areas of daily life. Road vehicles are also becoming more technologically advanced in terms of
automation with a continuing trend toward fully autonomous vehicles (Fagnant &
Kockelman, 2015). There are high expectations placed on the technology, such as decreasing
the number of road fatalities, reducing congestion, providing individual motorized mobility
solutions to people currently not allowed or not able to drive, and to enable users to engage
in other activities while driving (Anderson et al., 2014; Fagnant & Kockelman, 2015; Litman,
2014). Certain user groups might benefit more than others mainly depending on regular
time spent travelling. This is especially the case for people, such as commuters, who routinely
make long trips by car, have a limited time budget and hence mostly a high
willingness-to-pay (WTP) for saving travel time.
Commuting trips make up only a third of all trips in Germany but play a crucial role in road
traffic as they determine peak travel demand (DLR & Infas, 2008). In the past years,
commuting trips in Germany remained unchanged in terms of trip length (57% are shorter
than 10km), but increased slightly in terms of trip duration (22% take 30 to 60 min.; 4%
increase) suggesting that more commuters are stuck in congestion on the way to and from
work (Bundesamt, 2014). Heavy traffic conditions at peak hours suggest that extensive
commuting is often felt to be an exhausting and tedious task. A recent study on the
relationship between mode choice and commuting stress found that car drivers have the
highest stress levels compared to users of other modes. Furthermore, time consumption was
among the most important subjective stressors for commuters driving on a daily basis
(Legrain, Eluru, & El-Geneidy, 2015). Hence, an important benefit of having the opportunity
to ride autonomously for commuters might be that they can compensate time consumption
for commuting by using the time in a more efficient or more pleasurable way (Fraedrich,
Cyganski, Wolf, & Lenz, 2016; Trommer et al., 2016).
The range of activities that can be performed during a trip depends, however, on the degree
of automation of the vehicle. Referring to the definition given by the Society of Automotive
Engineers (SAE, 2014), only Level 4 (the system in charge, some driving use-cases) and Level
5 (the system in charge, all driving use-cases) achieve the degree of independence from
driving tasks that allow drivers to completely dedicate their attention to alternative activities.
Thus, this study deals only with those levels of automation.
High-level automation will also enable new mobility services such as vehicles on demand
either as individual ´autonomous carsharing´ service (ACS) similar to today’s carsharing and
taxi services or as ´autonomous ride sharing´ (ARS), when pooling different trips together
similar to uberPOOL
1
. ARS services are expected to exhibit lower costs per mile at somewhat
higher waiting times compared to ACS (Kröger & Kickhöfer, 2017). These services could
complement traditional public transport (e.g. solving the first/last mile problem) or act as a
substitute where it is deficient today (BCG, 2016; Ohnemus & Perl, 2016; Yap, Correia, &
1
Services like uberPOOL (https://www.uber.com/nyc-riders/products/uberpool) do not exist in today´s
mobility market in Germany
Working Paper
4
Arem, 2015). From a user perspective these services could allow true door-to-door trips for
individuals not having access to a car today (Burns, Jordon, & Scarborough, 2013).
In summary, it can be expected that the car be it privately owned or a shared vehicle will
become more attractive and, at the same time, available to broader user groups. This would
lead to rebound effects, resulting in more vehicles on the road, more congestion and/or more
vehicle miles traveled (Bahamonde-Birke, Kickhöfer, Heinrichs, & Kuhnimhof, 2017; Gruel &
Standford, 2016; OECD/ITF&CPB, 2015). Predicting changes in travel behavior and the traffic
situation today is hard to do, but is more and more relevant in light of uncertainties about the
future of mobility against the background of urbanization, demographic trends, and
environmental challenges.
The aim of this paper is to analyze how autonomous driving may change mode choices for
commuting trips. For this purpose, two different concepts of autonomous driving are
considered. The first use case is privately owned autonomous vehicles (AVs) able to drive
autonomously but with the option of switching off the autopilot. The second use case is
shared autonomous vehicle (SAVs), which combines (Uber-like) the concepts of taxi and
carsharing, where people can use a vehicle on demand. The results of the study should
provide empirical insights on future modal choice preferences for commuting trips.
LITERATURE REVIEW
The concept of value of travel time savings (VTTS) plays a crucial role in theoretical and
empirical literature in transportation research. In accordance with the microeconomic theory,
individuals are supposed to take transportation decisions under the assumption that the daily
time budget is constrained. Hence, people choose whether they spend their time in one
activity compared to another or how much are they willing to pay to save the time spent in
one particular activity (Hensher, 2011). The subjective VTTS can be defined, therefore, as the
willingness-to-pay to reduce the travel time (Jara-Diaz, 2000). VTTS usually depends on trip
purpose and trip length and differs between modes of transportation. Studies on VTTS
estimated higher values for commuting trips than for leisure or shopping trips (Abrantes &
Wardman, 2011; Shires & Jong, 2009). Also, the VTTS for commuting by car are in some
studies lower but in other higher than for public transportation and car passengers tend to
have a lower VTTS compared to car drivers (Abrantes & Wardman, 2011; Mackie et al., 2003;
Shires & Jong, 2009). Furthermore, various empirical studies found that the VTTS of business
travelers and commuters is higher in congestion than in free-flowing traffic (Abrantes &
Wardman, 2011; Hensher, 2011; Rizzi, Limonado, & Steimetz, 2012). This suggests that even
lower levels of automation might provide benefits for car users, for instance by enabling
automated stop-and-go functions in dense traffic. Furthermore, it can be assumed that
autonomous driving may potentially reduce VTTS for commuting trips in terms of perceiving
the travel time less negatively.
Lately, a significant body of literature has addressed the possible impact of AVs on travel
behavior (Childress, Nichols, Charlton, & Coe, 2015; Gucwa, 2014; Trommer et al., 2016).
However, given the lack of empirical studies, potential reductions in VTTS are usually
Working Paper
5
considered on the basis of plausible assumptions.
Filling this research gap is, however, a difficult task, as AVs are not currently available, so
there is no existing behavioral data. Alternatively, it is possible to rely on stated preferences,
but, in this case, respondents’ lack of experience can affect the reliability of the results. Thus,
it is advisable to center the analysis on high-level features, while acknowledging the
limitations of the technique. Hence, while attempting a detailed analysis of ground-breaking
mobility options may prove difficult, focusing the analysis on potential reductions in the VTTS
appears as a more plausible task for respondents. As far as we know, only a few
stated-choice studies addressing this topic have been conducted to date.
Yap et al. (Yap et al., 2015) address the use case of SAVs as egress transport for first/last mile
trips in multimodal train trips considering time and costs for the trip as well as sharing levels
(carsharing and ridesharing). The results of the study suggest that first/last mile AVs can be
attractive, especially for first class train travelers. Furthermore, the sensitivity of users for
in-vehicle time is higher in autonomous compared to manually driven vehicles, resulting in
higher VTTS, which the authors attribute to attitudinal and perceptual concerns toward the
technology. Along these lines, the results by Winter et al. (Winter, Cats, Martens, & Arem,
2016) show strong differences between early and late adopters (with a clear preference for
SAVs in the early adopters group) in the context of modal-choice, while including an SAV
alternative. The results of Krueger et al. (Krueger, Rashidi, & Rose, 2016) on the adoption of
SAVs show a similar trend. The authors found a strong impact of service attributed including
travel time, waiting time and fares as well as significant effects of individual-specific
charactersistics, such as age and individuals modality style on mode choices.
All three studies, while providing initial empirical insights on user preferences regarding AVs,
focus on the introduction of SAVs as an alternative to current modes of transportation. In
doing so, they ignore non-motorized alternatives, such as walking or cycling, but also the
option of privately owned AVs. Hence, we cannot gain from these studies insights on the
willingness to use privately owned AVs compared to SAVs.
Some recent studies also address user preferences toward privately owned AVs. However,
these studies focus on the impact of autonomous driving on car ownership or possession of
a public transportation pass. Becker & Axhausen (Becker & Axhausen, 2017) used a stated
choices approach to assess the impact of SAVs and privately owned AVs on mode choices.
Their pilot study with 62 participants suggests a decrease in car ownership rate by
introducing autonomous driving, especially as a sharing service. Another stated-choice study
on impact of AVs for commuting trips found that, besides cost, various attitudinal variables,
such as technology interest and enjoyment of driving, influence the user preferences toward
the technology (Haboucha, Ishaq, & Shiftan, 2017). However, the study focuses more on
long-term choice decisions than on influencing factors on trip mode choices.
In summary, we did not find any study focusing on the evolution of the VTTS related to the
introduction of AVs, nor studies addressing both privately owned AVs and SAVs
simultaneously, and compared to all other relevant modes of transportation.
Working Paper
6
METHODS
Study Design
In order to address the research questions, an online survey was conducted using a
questionnaire with the following structure: questions on existing mobility behavior, questions
on the commuting trip the person usually takes (i.e. reference trip), short introduction to the
concept of autonomous driving, a discrete choice experiment (DCE), questions on willingness
to purchase and pay for AVs as well as socio-demographics.
The study design was based on an earlier methodological approach which combines revealed
and stated preference data (Axhausen, 2014; Rose, Bliemer, Hensher, & Collins, 2005). In the
revealed preference part of the questionnaire, the respondents were asked to describe a
recent trip. In the stated choice experiments, hypothetical mode choice situations for the
same trip were constructed using the individual trip length of the respondents. In each choice
situation, the time and the cost for the trip were reduced or increased around reference
values using estimated average speeds and cost for each mode of transportation. The choice
experiment consisted of eight choice situations in which the respondents had to choose
between one of the following five transportation options: walk, bike, public transportation,
privately owned AV and an SAV. The SAV was called “driverless taxi” in order to provide a
better understanding of the concept to the participants. The attributes and their levels used
in the experiments are summarized in Table 1. It was assumed that an AV drives up to users,
drops them off and finds a parking spot by itself. Hence, access and egress time for the
autonomous vehicles were excluded as attributes and waiting time was considered.
Working Paper
7
Table 1: Attributes and attributes´ levels
Transport
mode
Attribute
Levels
Walk
Time
-30%|-10%|+20% reference Time [Speed: 4.9 km/h]
Bike
Time
-30%|-10%|+20% reference Time [Speed: 15 km/h]
Access Time
2 Min. | 5 Min.
Autonomous
vehicle (AV)
Time
-30%|-10%|+20% reference Time [Speed: between 26-68
km/h, distance dependant]
Waiting Time
2 Min. | 5 Min. | 10 Min.
Cost
-30%|-10%|+20% current costs [0.20 euro ct./km]
Shared
autonomous
vehicle (SAV)
Time
-30%|-10%|+20% reference time [Speed: between 26-68
km/h, distance dependant]
Waiting Time
2 Min. | 5 Min. | 10 Min.
Other
passengers
alone / other passengers
Cost
-30%|-10%|+20% reference costs “alone” [0.20 euro/km]
-30%|-10%|+20% reference costs “other passangers” [0.20
euro/km]
Public
Transportation
Time
-30%|-10%|+20% reference time [Speed: between 18-51
km/h, distance dependant]
Access Time
2 Min. | 5 Min. | 10 Min.
Waiting Time
2 Min. | 5 Min. | 10 Min.
Cost
-30%|-10%|+20% current costs [between 1.5 and 6 euros,
distance dependant]
In order to present realistic alternatives to the study participants, we used ´average speeds´
and ´cost per transportation mode´ for the German case. Average speeds were estimated
using the German National Household Travel Survey from 2008, called MiD 2008 (DLR &
Infas, 2008). The costs per kilometer for the private car were drawn from ADAC (ADAC,
2017). Only fuel and maintenance cost were taken into account. Cost related to the purchase
of the vehicle or parking cost were not considered. The kilometer price for the shared
autonomous vehicles followed existing analysis (Kröger & Kickhöfer, 2017). The cost for
public transportation was drawn from existing rates for public transportation systems in
Germany. We used fixed cost for different distance classes with a minimal price of 1.50 euro.
Season, annual or student tickets for public transportation were not considered.
In order to enhance the data quality of the experiments by maximizing the information
obtained from each choice situation, we created a Bayesian efficient design using the
software Ngene (ChoiceMetrics, 2012). Efficient design is recommendable when some initial
Working Paper
8
information about the value of the parameters is available prior to the field test, as it can
improve the design significantly and reduce the standard error (Bliemer & Rose, 2006). In our
study, the prior values for the estimation of the efficient design were drawn from a pilot study
with 30 respondents. The design was optimized for short and medium/long trips in order to
consider the effect of trip distance on trip and mode related attributes.
Introduction of the concept of autonomous driving
The two concepts of autonomous driving privately owned and shared AVs were presented to
the study participants in two short animated videos before the choice experiment. In the first
video the main character, Ms. Schmidt, calls her vehicle using an app on her phone, rides to
her pre-programmed destination, gets out of the car as she arrives and the vehicle drives
further autonomously and parks itself. In the second video, the concept of an SAV is
introduced. It is shown that one can order the vehicle, ride autonomously to one’s destination,
get out of the car and the vehicle drives on to collect its next passenger(s). The main
difference between the two introduced concepts was that, in the privately owned vehicle,
there is an option to switch off the autopilot. In the SAV there were no steering wheel and
brakes, it could not be driven manually. The two concepts were presented as neutrally as
possible (without using evaluative adjectives) in order to influence the preferences toward
autonomous driving as little as possible.
To find out if respondents prefer to drive their hypothetical privately owned vehicles
autonomously or manually, we added an additional question with a Likert scale related to this
preference after the choice experiment. Based on the responses, two dichotomous variables
were created which indicate whether they prefer to use their privately owned vehicles
autonomously or use them manually.
Implementation and Sample
For the online implementation of the questionnaire including the choice experiment the
software Sawtooth was used. Survey participants were recruited using a professional panel
service. A sample of 485 respondents representative for Germany by age and gender was
recruited. The sample included car users as well as non-car users and was limited to
participants older than 18. The duration of filling in the online survey was 13 minutes on
average. The respondents were randomly selected to provide information about one of three
different trip types - commuting trips, shopping trips and leisure trips. However, in this paper
a reduced sample size of 172 respondents was used since the rest of the sample reported
other trips than commuting.
A comparison between the reported commuting trips of our sample and commuting trips
from the German national travel survey MiD 2008 (DLR & Infas, 2008) shows that the key
parameters are largely similar (see Table 2). A critical point is the overrepresented public
transport use and by contrast, the underrepresented car use in our sample. The
mode-specific distances and times of commuting trips fit quite well. However, using trip
length and trip duration as reference parameters of the presented choice experiments the
existing data seem to be suitable.
Working Paper
9
Table 2: Comparison of the commuting trips between the German National Travel
Survey MiD 2008 (DLR & Infas, 2008) and the study sample
Bicycle
Car
Public
Transport
Mean
German National Travel Survey
Modal split (in %)
10
70
12
-
Commuting time (in min.)
15
26
53
27
Commuting distance (in km)
3.5
20.0
25.8
17.7
Study sample
Modal split (in %)
8
60
23
-
Commuting time (in min.)
14
24
46
27
Commuting distance (in km)
3.8
19.7
25.2
18.1
Analysis method
The most common alternatives in mode choice with multiple alternatives are the multinomial
logit (MNL) and the more advanced mixed logit (ML) models (Hensher & Greene, 2002). The
MNL model developed and described by McFadden (McFadden, 1974) estimates the
probability of each individual n selecting alternative i. Here it is assumed that n assigns a
given utility to every alternative i in the sampling, opting for the alternative that maximizes
the expected utility. Asuming additive linearity, the expected utility is given by the following
expression:
      (1)
Xn,i is a vector of explanatory variables including attributes of the alternatives as well as
socio-economic characteristics of the respondent, and are parameters to be estimated.
The error term εn,i repreesents a stochastic component, accounting for all relevant attributes
that are ignored by the modeler. An MNL imposes the condition that εn,i follows an
independent and identically (iid) extreme value type 1 distribution (McFadden, 1974).
However (and because of the restriction imposed upon the distribution of the stochastic
elements), the MNL does not allow considering heteregoneity among respondents nor
capturing the panel nature of our data. Thus, we rely on an ML to relax the assumptions that
the coefficients are the same for all individuals (Algers, Bergström, Dahlberg, & Dillen, 1998;
Train, 2002) and to allow correlation across choice situations (Hensher & Greene, 2002;
Revelt & Train, 1997). The utility function of an ML with panel data extends equation (2) as
followed:
        (2)
Here, the coefficient vector from equation (1) is expressed as . In this framework, bi
accounts for the population mean and is a random term following a distribution to be
established by the analysis with a given mean (normally zero) and denstity to be estimated.
This allows accounting for different valuations of Xn,i across individuals. t represents the
different choice situations a given individual n is confronted with, and therefore is
not assumed to vary across different choice sitatuions t, taking the panel effect into account
Working Paper
10
(i.e. the valuation of the attibutes remain constant for all observations associated with the
same individual). The ML probabilities of choosing given alternative i is, consequently, a
weighted mean of the MNL probabilities at a specific , weighted over thedistribution of .
   (3)
In (3) the choice probability  respresents the MNL probabilities for a given value of. Due
to the fact that an individual is faced with t choice situations, the probability of observing a
given sequence of choices is given by the following expression:
  


 (4)
Model specification
To obtain the final model specification, an iterative procedure was used. In the first step of
the analysis, an MNL was estimated only considering time and cost parameters. Afterwards,
socio-economic variables were introduced (solely siginificant socioeconomic variables were
finally part of the models). The final specification of the model considers the following
explanatory variables:
TTi: travel time of mode i (in minutes)
TCi: travel cost of mode i (in €)
SR: dummy for shared ride for driverless taxi
MAN: dummy for individual who prefers driving PAV manually
AUT: dummy for individual who prefers driving PAV autonomously
LH: dummy for license holder
AGEmiddle: dummy for middle aged individual (between 30 and 50 years old)
ATi: access and egress time for mode i (in minutes)
WTi: waiting time for mode i (in minutes)
INC: dummy for income class (low: up to 1.500 euros, middle: 1.500-3.000 euros, high:
more than 3.000 euros)
MALE: dummy for male gender
All explanatory variables are assumed to have a linear additive impact on the utility functions,
although not all of them affect the utility of all alternatives. Furthermore, it is assumed that
the alternative-specific constants (ASC) and the valuation of the generalized travel time (see
below) exhibit stochastic variations across individuals. The distribution of the associated
with the ASC and the generalized travel time is assumed to be normally distributed. The β
parameter associated with the cost of the alternatives is assumed to exhibit variation among
income classes.
Working Paper
11
In order to consider a decreasing marginal utility of time and costs on mode choices, we use
a Box-Cox transformation (Box & Cox, 1964). From a behavioral standpoint, this might -
especially in the case of commuting trips provide important insights on time perception and
VTTS depending on travel distance. The considered transformations are depicted in equation
(5).
,
( ) 1
Time
i Acc i Wait i
Time i
Time
TT AT WT

 
and
Cos
Cos
Cos
( ) 1
t
i
t
t
TC
(5)
Here, the expression considered in association with the time parameter represents the
generalized travel time, which takes into account that access and waiting time are perceived
differently from in-vehicle travel time. Here,  and  are also parameters to be
estimated, which also exhibit variability across individuals. However, in contrast to , the
distribution of  and  is considered to be uniform, in order to avoid problems with
negative values inside the Box-Cox transformation.
Finally, two ML models were estimated; one of those did not consider non-linearity, whereas
the other one considered the Box-Cox transformation. As previously mentioned, parameter
variability across individuals were only considered for time-related variables (i.e., travel time,
access and egress time and waiting time) and the ASCs. The estimation of the models was
preformed using PythonBiogeme (Bielaire, 2003). The distributions of the random
parameters were simulated by using 5,000 MLHS draws (Hess, Train, & Polak, 2006).
RESULTS
Estimated model coefficients
The results of the two final estimated ML models are summarized in Table 3. In general, the
coefficients exhibit the expected signs and plausible values. We obtain a significantly better
model fit by modeling possible non-linearity for the time and cost parameters (χ2 (2,
N=172)=9.65, p<.01). Hence, our results confirm the existence of decreasing marginal
utilities.
Working Paper
12
Table 3: Results of the two mixed logit model estimation
Model 1: Mixed logit
Model 2: Mixed logit with a
Box-Cox transformation for
time and cost
Coefficient
Estimated
Value
t-value
Estimated
Value
t-value
ASCPED
11.9
(4.02)
14.6
(4.75)
ASCBIKE
4.42
(4.62)
8.25
(4.32)
ASCPT
-3.39
(-3.27)
-2.68
(-2.07)
ASCSAV
-1.74
(-2.89)
-1.62
(-2.29)
η_PED
0.857
(0.71)
-0.372
(-0.35)
η_BIKE
3.64
(5.86)
3.53
(5.73)
η_PT
2.81
(3.35)
2.25
(2.92)
η_AV
1.38
(2.88)
-0.559
(-0.82)
η_SAV
-1.65
(-4.01)
-1.83
(-5.08)
ßTIME_PED
-0.423
(-4.56)
-1.31
(-1.87)
η_TIME_PED
-0.168
(-4.34)
-0.292
(-2.34)
ßTIME_BIKE
-0.314
(-7.51)
-1.35
(-1.90)
η_TIME_BIKE
-0.116
(-5.51)
-0.394
(-2.45)
ßTIME_PT
-0.0825
(-3.60)
-0.402
(-1.62)
η_TIME_PT
0.0703
(4.19)
0.254
(2.12)
ßTIME_AV_AUTONOM
-0.0784
(-3.69)
-0.307
(-1.65)
η_TIME_AV_AUTONOM
0.062
(2.48)
0.213
(2.17)
ßTIME_AV_MANUAL
-0.114
(-5.84)
-0.442
(-1.93)
η_TIME_AV_MANUAL
-0.0355
(-1.66)
-0.109
(-1.76)
ßTIME_SAV
-0.102
(-4.68)
-0.403
(-1.84)
η_TIME_SAV
0.0183
(0.80)
0.0324
(0.51)
ßWAIT (uniform-bottom)
1.08
(3.83)
1.01
(4.05)
η_WAIT (uniform-top)
3.28
(3.82)*
2.12
(4.45)*
ßACC
1.08
(3.22)
1.07
(3.97)
ßCOST_LOW_INC
-1.14
(-5.72)
-1.52
(-4.49)
ßCOST_MID_INC
-0.947
(-6.1)
-1.24
(-3.54)
ßCOST_HIGH_INC
-0.543
(-5.61)
-0.79
(-3.24)
ßSHARED
0.0191
(0.07)
-0.033
(-0.13)
λCOST
-
-
0.787
(5.89)
λTIME
-
-
0.566
(3.50)
ßPT_CARD
1.43
(1.71)
1.98
(2.54)
ßLICENCE_PED
-4.74
(-2.22)
-4.6
(-2.91)
ßMID_AGE_PED
-4.14
(-2.70)
-4.11
(-3.62)
ßMID_AGE_BIKE
-3.27
(-3.09)
-3.62
(-3.40)
Model Fit
Log-likelihood (final)
Estimated Parameters
Observations
-948.011
32
1376
-943.187
34
1376
*The t-values are referred to the bottom level of the uniform distribution.
Working Paper
13
Overall, the results show that cost and travel time elements influence mode choices
significantly, both having an expected negative impact. The coefficients in Model 2 are higher
than in Model 1 but have similar relations to eachother, suggesting stable tendencies.
The generalized time coefficients show differences between the modes. Travel time in
privately owned AVs is perceived less negatively by people using the automation function on
commuting trips compared to people driving manually. The accros-population variability of
the estimated coefficients suggests a wider heterogeneity among driving AVs automatically
than manually. Also, riding autonomously to work is perceived less negatively than the travel
time of any of the other available motorized alternatives. However, the differences are not
statistically significant.
When considering the ASCs, the general preference for SAVs is significantly lower compared
to privately owned vehicles; however the mode is more attractive than public transportation.
At the same time, looking at the travel time coefficients suggests that riding autonomously in
an SAV is perceived less negatively than driving, but is less attractive than riding
autonomously with a privately owned vehicle.
However, a comparison between the modes is only possible when considering all time
elements, including waiting and access/egress time. The coefficients for these two time
elements were estimated in relation to in-vehicle time. While there are no major differences
between acces and in-vehicle travel time (access tiem is perceived as slightly more negative),
waiting time is perceived 2.12 to 3.28 times more negatively (depending on the model) than
the in-vehicle time.
Furthermore, as expected, there is a relationship between cost sensitivity and household
income. People with low income are more cost-sensitive, perceiving travel cost more
negativily than people with middle or high income. This is reflected in the WTP differences
described in the following section.
The analysis of the perception of autonomous carsharing compared to autonomous
ridesharing (represented throught ßSHARED) did not provide any statiscally significant evidence
on whether people would prefer to share a ride with others or to ride alone in an SAV. This
suggests a smaller role of the sharing aspect compared to other factors.
Regarding the impact of socio-demographic factors, we included in the final model only the
variables found to exhibit a stattiscally significant effect. We found no significant effect of
gender on mode preferences in the final estimations. Regarding age, the analysis shows that
middle-aged people (between 30 and 50 years old) are less inclined to walk or cycle to work
than younger or older people. Possession of a public transportation pass influences
preferences for that mode positively. Furthermore, people who possess a driver’s license are
less inclined to walk to work. We did not find any socio-economic varaibles which were
directly related to preferences toward autonomous vehicles.
Estimation of VTTS
As previously mentioned the main objective of the analysis is to establish the differences
among the valuation of the travel time savings depending on transportation mode, when
AVs are available. This allows us to establish to which extent relieving the users from the
driving task may impact the time perception.
Working Paper
14
Establishing the VTTS is straightforward for Model 1, as we consider constant marginal
utilities of both travel time and costs, so that the VTTS can be established in accordance with
the following expression:
,
,
i
Time i
i
iCost n
i
U
TT
VTTS U
TC

(6)
For Model 2, considering decreasing marginal utilities for both travel time and cost, the VTTS
depends on the actual travel time and cost experienced by the user, as in the following
expression:
Cos
1
,
1
,
()
()
Time
t
i
Time i i Acc i Wait i
i
iCost n i
i
U
TT AT WT
TT
VTTS UTC
TC
 
 

(7)
Therefore, it is only possible to calculate an average for the considered population.
Furthermore, as the marginal utility of the price depends on the actual cost, the VTTS would
exhibit slight variation (<5% in our case) depending on alternative used as reference. In this
work, we have considered the marginal utility of the cost of SAVs as the reference to establish
the VTTS. The estimated values are summarized in Table 3.
Table 4: Estimated VTTS for different modes of transportation and income classes
(in euro/hour)
Walk
Bike
Public
transport
AV
autonomously
AV
manually
SAV
Model 1: Mixed logit
Low income
22.26
16.53
4.34
4.13
6.00
5.37
Middle income
26.80
19.89
5.23
4.97
7.22
6.46
High income
46.74
34.70
9.12
8.66
12.60
11.27
Model 2: Mixed logit with a Box-Cox transformation for time and cost
Low income
8.88
13.41
3.93
3.74
5.39
4.85
Middle Income
10.88
16.44
4.81
4.59
6.60
5.94
High Income
17.08
25.88
7.56
7.20
10.36
9.32
The results for the VTTS reflect the results from the estimations presented above. People with
a high income have a higher willingness-to-pay for saving commuting travel time. Here,
again, the VTTS for people who prefer autonomously driving privately owned AVs is lower
than the VTTS of people driving manually by 31% in both models. It reflects the perceived
benefits of relieving the user from driving tasks and allowing them to dedicate their attention
to activities deemed as more meaningful. The VTTS for driving autonomously is in the range
of VTTS for in-vehicle time in public transportation, suggesting a similar perception for both
modes of transportation. However, it does not include waiting and access/egress time, which
can be, as estimated above, up to 2 or 3 times more negative than in-vehicle time (this
phenomenon negatively affects the perception of public transportation). At the same time,
Working Paper
15
the VTTS for SAVs is slightly higher than autonomously driven vehicles and public
transportation, but still 10% lower than for driving a car by oneself. Hence, using an SAV
may be deemed more attractive than driving manually to work (although relying on SAVs
may also involve waiting time).
CONCLUSIONS
The main aim of this study was to analyze how autonomous driving may affect the subjective
value of travel time savings for commuting trips. For this purpose, a discrete choice
experiment was conducted and the data were analyzed using a mixed logit model.
First, the results provide empirical evidence supporting the assumption that autonomous
driving will potentially reduce the VTTS for commuting trips, i.e. it will be an attractive
function for people making regular commuting trips. Moreover, the VTTS for two different
possible uses of autonomous driving, namely privately owned AVs and SAVs, were estimated
for different income classes and contrasted with alternative modes of transportation. Our
results suggest that driving autonomously leads to a reduction of 31% in the VTTS compared
with driving manually, and is perceived similarly to the VTTS of in-vehicle time in public
transportation (waiting and access/egress time is perceived more negatively in public
transportation).
Second, when considering the preferences toward SAVs, we found that travel time spent in
SAVs is perceived less negatively than driving manually by 10%. However, riding
autonomously privately owned AVs seems to be more attractive than using SAVs. In general,
the preference for using privately owned vehicle in the sample seems to be higher than using
shared vehicles; at least for regular commuting trips. Even though the VTTS in SAVs seems to
be a little higher than the in-vehicle VTTS for public transportation, it does not include larger
waiting and access/egress time associated with public transportation (and the fact that the
travel time in public transportation is usually greater than by car). This suggests potential for
SAVs as an alternative (or complementary service) for public transportation.
Regarding different user perceptions towards autonomous ridesharing compared to
autonomous carsharing, our study does not offer conclusive results. However, users´
concerns about sharing a ride with strangers are possible. Thus, attitudes towards sharing a
ride have to be considered in future works, for instance using a sample of people with
ridesharing experience, such as users of uberPOOL or of private-organized ridesharing.
The main limitation of the study is related to possible hypothetical bias as AVs are not
currently available. Therefore, providing realistic answers may be difficult for the respondents,
as they do not have direct experience with the technology. Therefore, while acknowledging
the limitations of the technique, we have centered the analysis on a high-level feature, the
VTTS, which may be easier for the respondents to internalize.
In all, the study provides important empirical evidence and insights into how autonomous
driving might affect mode choices and valuation of travel time for commuting trips. This
study thus lays groundwork on the possible impacts of introducing AVs on the valuation of
travel time, which future research can build upon. Along the same lines, the study provides
Working Paper
16
empirical evidence sustaining the reduction of the VTTS considered by many authors in
simulations exercises.
Future research should focus on other relevant determinants of mode choices, and also on
understanding the perception of in-vehicle time for autonomous driving, which has not been
covered in this study. In any case, caution is required, as respondents may be overwhelmed
when confronted with groundbreaking technologies they are not familiar with. Thus, the
analyst should focus their efforts on aspects the respondents can deal with. Another avenue
for future research may be understanding determinants behind user preferences and
perception. Hence, further work on users attitudes and needs as well as perceived individual
benefits of automation might be crucial in understanding commuters’ decision-making
processes.
REFERENCES
Abrantes, P. A. L., & Wardman, M. R. (2011). Meta-Analysis of UK Values of Travel Time: An
Update. Transportation Research Part A: Policy and Practice, 45(1), 1-17.
ADAC. (2017). ADAC Autokosten Frühjahr/Sommer 2017 Retrieved from München,
Germany https://www.adac.de/_mmm/pdf/autokostenuebersicht_47085.pdf
Algers, S., Bergström, P., Dahlberg, M., & Dillen, J. L. (1998). Mixed Logit Estimation of the
Value of Travel Time. Retrieved from
Anderson, J. M., Kalra, N., Stanley, K. D., Sorensen, P., Samaras, C., & Oluwatola, O. (2014).
Autonomous Vehicle Technology - A Guide for Policymakers. Retrieved from
Axhausen, K. E., Ilka; Glemser, Axel; Hess, Stephane; Jödden, Christian; Nagel, Kai; Sauer,
Andreas; Weis, Claude. (2014). Ermittlung von Bewertungsansätzen für Reisezeiten und
Zuverlässigkeit auf der Basis eines Modells für modale Verlagerungen im nicht-gewerblichen
und gewerblichen Personenverkehr für die Bundesverkehrswegeplanung. Retrieved from
Bahamonde-Birke, F. J., Kickhöfer, B., Heinrichs, D., & Kuhnimhof, T. (2017, 21.-23. 06.2017).
A systemic view on autonomous vehicles: Policy aspects for a sustainable transportation
planning. Paper presented at the Kuhmo-Nectar Conference of the International
Transportation Economics Association, Barcelona, Spain.
BCG. (2016). Self-Driving Vehicles, Robo-Taxis, and the Urban Mobility Revolution. Retrieved
from
Becker, F., & Axhausen, K. W. (2017, May 17. 19. 2017). Predicting the use of automated
vehicles. [First results from the pilot survey]. Paper presented at the 17th Swiss Transport
Research Conference (STRC), Monte Verità / Ascona.
Bielaire, M. (2003, March, 19.-21.). A free package for the estimation of discrete choice
models. Paper presented at the 3rd Swiss Transportation Research Conference Ascona,
Switzerland.
Bliemer, M. C. J., & Rose, J. M. (2006, August, 16.-20.). Designing Stated Choice
Experiments: State-of-the-Art. Paper presented at the 11th International Conference on
Working Paper
17
Travel Behaviour Research, Kyoto, Japan.
Box, G. E. P., & Cox, D. R. (1964). An analysis of transformations. Journal of the Royal
Statistical Society Series B, 26(2), 211-252.
Bundesamt, S. (2014). Berufspendler: Infrastruktur wichtiger als Benzinpreis. STATmagazin:
Arbeitsmarkt.
Burns, L. D., Jordon, W. C., & Scarborough, B. A. (2013). Transforming personal mobility.
Retrieved from
Childress, S., Nichols, B., Charlton, B., & Coe, S. (2015, January, 11.-15.). Using an
activity-based model to explore possible impacts of automated vehicles. Paper presented at
the The 94th Annual Meeting of the Transportation Research Board Washington, USA.
ChoiceMetrics. (2012). Ngene 1.1.1 User Manual & Reference Guide Retrieved from
DLR, & Infas. (2008). Mobilität in Deutschland (MiD) 2008. Techn. rep. Retrieved from
Fagnant, D. J., & Kockelman, K. (2015). Preparing a nation for autonomous vehicles:
opportunities, barriers and policy recommendations. Transportation Research Part A 77,
167181.
Fraedrich, E., Cyganski, R., Wolf, I., & Lenz, B. (2016). User perspective on autonomous
driving. A Use-Case-Driven Study in Germany. Arbeitsberichte. Gerographisches Institut,
Humboldt-Universität zu Berlin, 187(187), 112.
Gruel, W., & Standford, J. M. (2016). Assessing the long-term effects of autonomous
vehicles: a speculative approach. Transportation Research Procedia, 13, 18-29.
Gucwa, M. (2014, June, 15.-17.). Mobility and Energy Impacts of Automated Cars. Paper
presented at the Automated Vehicle Symposium, Burlingame, CA, USA.
Haboucha, C. J., Ishaq, R., & Shiftan, Y. (2017). User preferences regardign autonomous
vehicles. Transportation Research Part C, 78, 37-49.
Hensher, D. A. (2011). Valuation of travel time savings. In A. d. Palma, R. Lindsey, E. Quinet, &
R. Vickerman (Eds.), A Handbook of Transport Economics: Edward Elgar Publishing Limited.
Hensher, D. A., & Greene, W. H. (2002, 10.06.2002). [The Mixed Logit Model: The State of
Practice ].
Hess, S., Train, K. E., & Polak, J. W. (2006). On the use of Modified Latin Hybercube Sampling
(MLHS) method in the estimation of a Mixed Logit Model for vehicle choice. Transportation
Research Part B, 40, 147-163.
Jara-Diaz, S. R. (2000). Allocation and valuation of travel-time savings In D. A. Hensher & K. J.
Button (Eds.), Handbook of Transport Modelling Elsevier Science Ltd.
Kröger, L., & Kickhöfer, B. (2017, June, 21.-22.). Autonomous car- and ride-sharing systems:
A simulation-based evaluation of various supply options for different regions. Paper
presented at the ITEA Annual Conference on Transportation Economics, Barcelona.
Krueger, R., Rashidi, T. H., & Rose, J. M. (2016). Adoption of Shared Autonomous Vehicles
A Stated Choice Analysis.
Working Paper
18
Legrain, A., Eluru, N., & El-Geneidy, A. M. (2015). Am stressed, must travel: The relationship
between mode choice and commuting stress. Transportation Research Part F: Traffic
Psychology and Behaviour, 34, 141-151. doi:https://doi.org/10.1016/j.trf.2015.08.001
Litman, T. (2014). Ready or waiting? Traffic Technology International.
Mackie, P., Wardman, M., Fowkes, A., Whelan, G., Nellthorp, J., & Bates, J. (2003). Values of
Travel Time Savings in the UK - Full Report. Retrieved from
www.its.leeds.ac.uk/working/downloads/VOTFull.pdf.
McFadden, D. L. (1974). Conditional Logit Analysis of Qualitative Choice Behavior In P.
Zarembka (Ed.), Frontiers in econometrics: Academic Press: New York.
OECD/ITF&CPB. (2015). Urban Mobility System Upgrade. How shared self-driving cars could
change city traffic. Retrieved from
Ohnemus, M., & Perl, A. (2016). Shared Autonomous Vehicles: Catalyst of new mobility for
the last mile? Built Environment 42(4), 589-602. doi:https://doi.org/10.2148/benv.42.4.589
Revelt, D., & Train, K. (1997). Mixed Logit with Repeated Choices: Households’ Choices of
Appliance Efficiency Level. Review of Economics and Statistics.
Rizzi, L. I., Limonado, J. P., & Steimetz, S. S. C. (2012). The impact of traffic images on travel
time valuation in stated-preference choice experiments. Transportmetrica.
Rose, J., Bliemer, M., Hensher, D., & Collins, A. (2005). Designing efficient stated choice
experiments involving respondent based reference alternatives. Working paper. Retrieved
from
SAE. (2014). SAE International Standard J3016.
Shires, J. D., & Jong, G. C. d. (2009). An International Meta-Analysis of Values of Travel Time
Savings Evaluation and Program Plannung 32(4), 315-325.
Train, K. E. (2002). Discrete Choice Methods with Simulation: Cambrige University Press.
Trommer, S., Kolarova, V., Fraedrich, E., Kröger, L., Kickhöfer, B., Kuhnimhof, T., . . . Phleps, P.
(2016). Autonomous Driving: The Impact of Vehicle Automation on Mobility Behaviour.
Retrieved from
http://www.ifmo.de/tl_files/publications_content/2016/ifmo_2016_Autonomous_Driving_2
035_en.pdf
Winter, K., Cats, O., Martens, K., & Arem, B. v. (2016). A Stated Choice Experiment on Mode
Choice in an Era of Free-Floating Carsharing and Shared Autonomous Vehicles.
Transportation Research Board 96th Annual Meeting.
Yap, M. D., Correia, G., & Arem, B. v. (2015). Valuation of Travel Attributes for Using
Automated Vehicles as Egress Transport of Multimodal Train Trips. Transportation Research
Procedia, 10, 462-471.
... Gurumurthy and Kockelman (2020) believe lower additional in-vehicle travel time of SAVs will increase WTU for these modes. The waiting time and delay were among the factors that most affected using SAVs Steck et al., 2018). Zhou et al. (2020) showed that people are generally sensitive to waiting time for using SAVs, especially those who regularly use shared services for commuting and have higher education levels. ...
... Travel time and costs, historically, are the main factors in choosing an automated shared transit mode (Gurumurthy and Kockelman, 2020;Krueger et al., 2016;Steck et al., 2018); therefore, along with similar reviewed studies, travel time has the highest importance. The main portion of participants (45 %) were willing to save $3 by adding only 5 min to their trip; however, fewer respondents (35 %) tended to maintain their minimum travel time of 10 min for their trips. ...
Article
Due to recent technological developments, both academia and the market have become interested in utilizing shared autonomous vehicles (SAVs). Gaining insight into preferences and priorities for using the next generation of shared mobility was the primary motivation for this study. The goal is to explore the willingness-to-pay (WTP) for a shared mobility service using a detailed user preference analysis for alternative automation types and incentive and policy options. The data was collected between March to May 2020 during the COVID-19 lockdown period in the United States via an online survey and analyzed by the Adaptive Choice-Based Conjoint Analysis (ACBC) method. The results of this study show that respondents distinctly preferred fully automated vehicles (SAVs) over shared human-driven vehicles (SHVs), tended to use a door-to-door service rather than a pooled service with a short walking distance, and were more interested in using advanced built-in amenities in the vehicle, not inclined to share their travel with others, and likely to pay less money to avoid potential delays. Comparing the results before and during the COVID-19 period showed that respondents afterward have markedly less willingness-to-share their trips with other riders compared to the survey analysis done before the COVID-19 period.
... Several scholars reveal the impact of AVs on various actors' travel behavior by using different methods, such as stated preference surveys, agent-based modeling, economic models, or empirical experiments (Hao and Yamamoto, 2017;Kopelias et al., 2020;Steck et al., 2018). In this section, several stakeholders connected to AVs are presented. ...
... AVs represent a potential transport mode that minimizes the value of travel time (VOT) and enables travelers to use their travel time more efficiently. The VOT of high-income travelers in AVs is less than in case of CTMs (Steck et al., 2018). Finally, all travelers can use AVs regardless of owning a driving license (Krueger et al., 2016). ...
Article
The availability of autonomous vehicles (AVs) on the market will create a novel situation for every stakeholder. A lot of research has been conducted on developing the AVs and the consequences of having AVs on the market, but comprehensive studies concerning various viewpoints on the introduction of AVs into the market can be scarcely found. This research is examining different actors' viewpoints on the acceptance of privately shared autonomous vehicles (PSAVs). Four groups of stakeholders are identified as the following: users, legislators, operators, and manufacturers. The multi-actor multi-criteria analysis (MAMCA) is used, where the analytical hierarchy process (AHP) and the parsimonious AHP (PAHP)methods are applied to evaluate each actor's objectives and criteria. From each group, a representative sample is collected, the actors evaluate a set of pairwise comparison matrices (PCMs), and their consistency is continuously checked. As a result, the objectives and the criteria are ranked and presented. The main finding of the objectives' analysis presents that the safety concerns receive the highest ranking (i.e., weight is 0.085), while the ease of use and the interoperability across the borders have the lowest rankings ((i.e., weight is 0.048, and 0.0515, respectively). The main finding of criteria's analysis shows that ease of use has the lowest ranking (i.e., weight is 0.0024), and the highest rank is related to the reduction in vehicle accidents caused by the malfunction (i.e., weight is 0.0495). Thus, it seems that safety related issues are the most important factor in accepting PSAV. The result of this study is useful for decision-makers and transport planners to form policies, regulations, and guidelines regarding the future implementations of PSAVs before their arrival to the market.
... Steck et al. [18] also conducted a pseudoexperiment in Germany considering five transport modes: walking, cycling, AV, SAV, and public transport. e selected attributes were trip time, access time, waiting time, cost, and whether the SAV trip would be made with other passengers. ...
... We calculated average speeds based on available data from a mobility study in Spain [20]. We obtained travel costs from studies about a standard vehicle's fixed and variable operating costs [21]. is design also considered previous examples such as those of Steck et al. [18] and Ashkrof et al. [14]. Table 2 presents the attribute levels used in the survey design. ...
Article
Full-text available
The possible future introduction of Autonomous Vehicles (AVs) and Shared Autonomous Vehicles (SAVs) raises questions about how they might affect the demand for transport and especially modal choice. In this research, a stated preference (SP) survey and a modelling process using Mixed Logit are proposed to simulate the future market share of AVs/SAVs and how their introduction into the system could change the modal choice, especially in relation to active and public transport modes. An efficient SP survey design has been developed based on the state-of-the-art information and carried out in 2020 among citizens of two medium-sized Southern European cities within a car-intensive region. The design considered different trip purposes (compulsory, leisure), different trip distances, and attributes not taken into account before, such as comfort and the physical characteristics of the terrain for the active modes. The model results suggest that AVs and SAVs were the preferred transport modes for most respondents, accounting for more than 58% of the market share in the scenarios presented. Also, we detected some socioeconomic differences in the propensity to use this mode of transport showing that men living in high-income households and car users were more prone to use autonomous alternatives. The models allowed us to simulate different scenarios, such as experiencing higher costs for using the AV alternative. Policies imposing a higher cost for the AV alternative but lower costs and waiting times for the SAV and public transport alternatives could decrease the AV’s market share favouring more sustainable modes. The above scenario showed that achieving a more sustainable future mobility system considering AVs requires an in-depth transport demand knowledge and adequate transport policies.
... A number of authors, based on different types of research, estimate different values for VoT reduction: 41% compared with private cars [27], and 8-32% (8-14% for shared automated vehicles) compared with driving [28]. In some cases [29], it has been found that driving a private automated vehicle might reduce the VoT by 31% compared with driving manually, but travelling in a shared automated vehicle is perceived 10% less negatively than driving manually. Research [30][31][32] has shown the influence of socio-demographics, trip characteristics, and on-board activities on monetizing the value of time. ...
Article
Full-text available
New transport technologies, such as autonomous vehicles, are increasingly discussed in the debate on the transition to a sustainable urban future. Automated vehicles (AVs) are expected to reduce the value of travel time (VoT), allowing the use of time for other types of activities during travel, including working, reading, sleeping, entertainment, etc. Our study aims to provide empirical insights on future modal choice preferences for regular trips for Romanian citizens, using a sample of 309 respondents to a web survey on issues related to automated vehicles. Using multinomial logistic models (MNL), we analysed the relationship between three mode choices: regular car, private automated vehicle, and shared automated vehicle, along with the individual and household characteristics. In addition, we calculated the VoT for each mode choice based on the results of MNL analysis. Results showed that VoT is strongly influenced by travel cost and travel time, by socio-economic characteristics such as age, gender, and education, and has the lowest value for the shared AV compared with a regular car or a private AV. Future research may conduct comparable studies in European countries but also explore the opinions and perceptions of vulnerable road users on AVs and VoT.
... Furthermore, the assumption that VOT is constant for all trip purposes could be misleading, as survey-based studies suggest that it could vary depending on trip purpose, estimating VOT decrease at 30-40% for commuting trips [79] and zero for leisure trips [69]. Only one study implemented a differentiation of VOTsubject to the purpose of the trip, claiming VOT to be 100% of wage for business trips and 50% for leisure trips [29]. ...
Article
Full-text available
Advancement in the fields of electrification, automation, and digitalisation and emerging social trends are fuelling the transformation of road transport resulting in the introduction of various innovative mobility solutions. Yet the reaction of people to many of the new solutions is still vastly unknown. This creates an unprecedented quandary for transport planners who are requested to design future transport systems and create the related investment plans without fully validated models to base the assessment upon. As some evidence on citizens’ behaviour concerning new mobility solutions starts to be progressively made available, first attempts to update the existing models begin to emerge. Nevertheless, a lot more is needed as some of the transpiring mobility solutions have not yet reached the market, making the corresponding behaviour changes imponderable. In this context, the main purpose of this paper is to provide a review on how travel behaviour changes linked to the deployment of new mobility solutions have been considered in travel demand models. The new mobility solutions studied include carsharing, dynamic ridesharing, micromobility sharing services, and personal and shared autonomous vehicles. An overview and comparison of relevant studies implementing activity or trip-based demand models and other methodologies are presented. The analysis shows that the results of the different studies heavily depend on the extent to which behavioural changes are considered. The results of the review thus point to the need for holistic demand models that carefully mimic the urban reality with everything it has to offer and account for the importance of individual traits in the decision-making processes. Such models need an in-depth understanding of the microscopic mechanisms leading to the travel behaviour shifts linked to the most innovative mobility solutions. To achieve this level of detail, mobility living labs and their real-life experiments and experience with citizens, which are flourishing in Europe, are suggested to play a crucial role in the years to come.
Article
In this work we are concerned with how the introduction of connected and automated mobility (CAM) will influence liveability in cities. We engaged with city and transport planners from both Europe and the U.S. and adopted a system dynamics approach to capturing the discussions and exploring potential outcomes. There are two aims in doing this: (1) to identify the concerns of city planners and how they differ from the traditional focus of transport researchers; but also (2) to develop a causal loop diagram (CLD) that can both explore the potential systemic effects of CAM and help to communicate those effects and the underlying mental models. Addressing these aims can inform policy design related to both CAM specifically and urban mobility more generally. In a change from previous related studies, we allowed the participants to establish their concept of liveability in cities and did not define a specific CAM scenario. This broad scope was critical in capturing the high-level view of what really matters to city stakeholders. We have established that a focus on a more holistic understanding of interactions related to sustainability is required rather than on specific transport modes or technology. A key insight that emerged was that quality of life (QoL) was the dominant concern of city planners, regardless of how it is achieved. The specifics of new services or technologies (such as CAM) are secondary concerns - which are important only insofar as they support the higher goal of improving QoL. As a result, we have produced a high level CLD that can be used as a starter for any future research in the area of CAM and liveability in cities and which may resonate better than previous CAM models have with city planners and policy makers—those who will ultimately play a key role in recommending and then implementing changes affecting QoL.
Article
Self-driving technology may lead to a paradigm shift for the transport industry with shared cars available to every-one. However, this vision has increasingly been challenged as too optimistic and unsubstantiated. In this study we explore societal impacts of using this technology for both cars and public transport and investigate differences depending on geography and trip purpose. Four scenarios were designed through workshops with 130 transport experts, modelled using a conventional four-step model for Stockholm, Sweden and evaluated in terms of changes to mode choice, number of trips and person kilometres. We find larger increases for non-commuting trips, i.e. service and leisure trips, than for commuting trips, questioning the view of the ‘productive work trip’ as self-driving technology’s main impact on society. As these trips are primarily made outside of rush hours, this may lead to a changed transport system. Geographic differences are substantial and heavily dependent on the cost model for car alternatives, even indicating a reduction in car travel in rural areas if private ownership would be replaced by shared cars. Furthermore, walking and cycling levels decreased in all scenarios while enhancing public transport using self-driving technology had a limited impact on ridership. These results show that the impacts of self-driving technology may have varied societal impacts even within a region and may lead to increased car travel, especially off-peak. These conclusions stress the need for policies that are sensitive to both geography and time.
Article
Smartphones can lower the disutility of waiting by increasing productivity and making time pass more pleasantly. We elicit the compensation required by subjects to wait for 30 minutes, alone in an empty room, under four different conditions that varied access to the subject’s smartphone. Compared to the treatment where subjects had full use of their phone, we find that they required 24% percent more to wait with the audio features of the phone remaining but the phone physically locked away, 48% percent more to wait with only an FM radio, and 79% percent more to wait in a quiet room. We find little correlation between a subject’s wages and her offers, emphasizing the importance of heterogeneity in the value of time that is based on context rather than income.
Article
This research monetizes the access benefits of making shared autonomous vehicles (SAVs) available to residents of Texas’ Dallas-Fort Worth metroplex in the U.S. Residents’ willingness to pay for SAV access under different fares and modes was estimated and compared across the region’s 5,386 traffic zones, with emphasis on those housing the regions’ most vulnerable or access-limited travelers. Assuming a $0.50/mi SAV fare, the average per-person-trip benefit is estimated to be $0.64 per trip. With $0.50/mi SAV fare, HV mode share will be reduced from 92.4% to 40.3%, while SAV will take 55.8% of the share. However, if HVs are then disallowed (removed from everyone s mode choice set) after $0.50/mile SAVs have been added, the average net impact is estimated to be −$0.31 per trip, across the metroplex. If HVs were to be replaced by access to SAVs, the impacts are positive, with region-wide average access benefits ranging from $0.16 to $0.33 per trip, depending on the SAV fare, and urban zones have a greater access benefit than rural zones with low SAV fare. Vulnerable populations and their neighborhoods were identified based on the share of persons living below the poverty level, income per capita, share of persons aged 65 years or older, those with disabilities, those owning no vehicle, and share of persons from a racial minority group. Results suggest that the access benefits of SAVs will be higher in locations/neighborhoods housing more vulnerable populations, but some vulnerabilities (e.g., those over age 65) results in lower levels of access improvement. Across those zones with highest shares of vulnerable persons, the range of differences in welfare impacts, from adding SAVs to travelers mode choice sets, widened as fares rose. As is true with many innovations, careful attention to disadvantaged groups and thoughtful policy (via smart contracting and SAV-user subsidies by public agencies, for example) can better ensure valuable access improvements for those with limited mobility and resources.
Article
Privately-owned automated vehicles (PAVs) can relocate themselves elsewhere after arriving at destinations, thereby inducing empty vehicle mile traveled (VMT) and more greenhouse gas emissions. This paper examines travelers’ preferences for PAV relocation, using stated preference surveys distributed in the Seattle and Kansas City regions in the U.S. Model results suggest that trip purpose, individual socio-economic and household characteristics, and local contexts influence PAV relocation decisions. Additionally, a willingness-to-relocate (WTR) metric is defined to represent how much time travelers would be willing to spend to relocate PAVs to save $1 in parking costs. The WTR estimates in both regions indicate the potential of induced VMT due to PAV parking relocation. Furthermore, travelers’ awareness of fuel/energy costs associated with PAV relocation does not necessarily relate to a lower WTR, depending on the region. To curb excessive empty VMT from PAV relocation, explicit disincentives, such as a VMT fee, may be needed.
Conference Paper
Full-text available
New forms of shared mobility such as Free-Floating Carsharing services and Shared Autonomous Vehicles have the potential to change urban travel behaviour. In this paper a stated choice experiment on mode choice among a sample of the Dutch urban population is presented, in which the particular features of free-floating carsharing and shared autonomous vehicles in comparison to private vehicles and public transportation are examined. The most explanatory and robust mode choice models were obtained by estimating nested logit models with two categories capturing vehicle automation or vehicle ownership, and a nested logit model with three categories capturing who is performing the driving task (the commuter, a human driver or an autonomous vehicle). Interpreted as mode preference, the alternative- specific constants of the utility functions reveal a strong impact of vehicle automation on mode choice: while early adopters of mobility trends show a clear preference for shared autonomous vehicles over all other modes, normal and late adopters show a clear aversion towards this mode. In terms of vehicle sharing, no preference of sequentially shared modes over a simultaneously shared bus could be determined. Participants currently not having access to carsharing services show a stronger preference towards free-floating carsharing than the early adopters subscribed to carsharing.
Technical Report
Full-text available
To analyse and quantify the possible impact of autonomous driving on mobility behaviour, we used an approach that combined both qualitative and quantitative methods. In the first phase of the project, we analysed the status quo in our selected countries and the expected developments in autonomous driving, and identified potential user segments and influencing areas related to autonomous driving which affect mobility behaviour. The insights were drawn from expert workshops, and from focus groups with potential users. In the second part of the study, using the results derived from the first phase of the project, we developed three scenarios to be examined with respect to the selected countries, which differed in the projected share of AVs within the vehicle fleet, as well as the way in which they are used as shared vehicles. Using a travel demand model, the impact of autonomous driving on mobility behaviour has been quantified.
Article
Full-text available
Among the first larger research activities that has dealt with autonomous driving from a wider perspective was the “Villa Ladenburg” project (2012-2015). Central aspects of the project included potential future users of autonomous vehicles and possible implications for the transport system. “Villa Ladenburg” was funded by the Daimler and Benz foundation, and convened an international network of renowned experts from various disciplines. The results of this work were published in a comprehensive compendium in German in 2015 (publicly available online: http://www.springer.com/de/book/9783662458532). An English version will be available around March 2016. Associated with the “Villa Ladenburg” project, an online survey with 1,000 respondents in Germany was conducted in April 2014. Its aim was to gather use-case-oriented mindsets concerning acceptance, mode choice, and time use related to expectations, desires, reservations, and fears concerning autonomous driving. Above all, we wanted to find out what the respondents imagine their mobility and transport behavior to be like in specific “what if…” scenarios. To what extent the respondents could actually imagine themselves in a future with autonomous vehicles is an important aspect in this early stage of technology development and implementation. Autonomous driving, in relation to its potential types of application and potential users, implicates changes on various levels, be it transport behavior, modifications in mindsets towards the use and ownership of cars, or altered use of time while traveling, thus indicating changes in terms of future activities. An interdisciplinary approach that includes perspectives from transport and mobility research, psychology, and social sciences is required to adequately address this topic, where still little is known and many questions remain unanswered. In the survey, we therefore applied a mix of methods to deal with these uncertainties. For example, we thought open-answer options in the form of free-text boxes a suitable addition to the standardized, quantitative questions, thus giving respondents an additional opportunity for more spontaneous reactions towards autonomous driving. At the same time, we put an emphasis on different use cases representing likely application scenarios of autonomous driving. The following results present initial, careful quantifications in a field, where many questions are still do be addressed. However, a focus on user perspectives onautonomous driving is crucial for a successful implementation of the technology into our transport system in the future.
Article
This study gains insight into individual motivations for choosing to own and use autonomous vehicles and develops a model for autonomous vehicle long-term choice decisions. A stated preference questionnaire is distributed to 721 individuals living across Israel and North America. Based on the characteristics of their current commutes, individuals are presented with various scenarios and asked to choose the car they would use for their commute. A vehicle choice model which includes three options is estimated: (1) Continue to commute using a regular car that you have in your possession. (2) Buy and shift to commuting using a privately-owned autonomous vehicle (PAV). (3) Shift to using a shared-autonomous vehicle (SAV), from a fleet of on-demand cars for your commute. A factor analysis determined five relevant latent variables describing the individuals’ attitudes: technology interest, environmental concern, enjoy driving, public transit attitude, and pro-AV sentiments. The effects that the characteristics of the individual and the autonomous vehicle have on use and acceptance are quantified through random utility models including logit kernel model taking into account panel effects.Currently, large overall hesitations towards autonomous vehicle adoption exist, with 44% of choice decisions remaining regular vehicles. Early AV adopters will likely be young, students, more educated, and spend more time in vehicles. Even if the SAV service were to be completely free, only 75% of individuals would currently be willing to use SAVs. The study also found various differences regarding the preferences of individuals in Israel and North America, namely that Israelis are overall more likely to shift to autonomous vehicles. Methods to encourage SAV use include increasing the costs for regular cars as well as educating the public about the benefits of shared autonomous vehicles.
Article
We are now witnessing the initial deployment of technology that promises to redefi ne what it means to travel by automobile and reshape the auto's future role in transportation networks. After decades of research, autonomous vehicles (AVs) are entering the mobility marketplace. AVs promise to make automobile travel safer by averting crashes, more equitable by providing mobility for the elderly, disabled, low-income, and non-drivers, and more effi cient by increasing road capacity, and allowing for driving time to be used for work or entertainment. A world of automated driving is likely to become a world where individual car ownership diminishes due to risk avoidance in the early adoption of this technology. Shared autonomous vehicles (SAVs) would retain door-to-door travel without the costs and congestion that accompany single occupant vehicles. Through their potential to connect the fi rst and last mile of trips in low-density areas, integrating SAVs with public transport systems could substantially increase synergies between autos and transit. Lowdensity land use could thus be shielded from climate and oil vulnerability as SAVs maintain accessibility to auto-dependent locations during times of climate and energy disruption.
Article
In the analysis of data it is often assumed that observations y1, y2, …, yn are independently normally distributed with constant variance and with expectations specified by a model linear in a set of parameters θ. In this paper we make the less restrictive assumption that such a normal, homoscedastic, linear model is appropriate after some suitable transformation has been applied to the y's. Inferences about the transformation and about the parameters of the linear model are made by computing the likelihood function and the relevant posterior distribution. The contributions of normality, homoscedasticity and additivity to the transformation are separated. The relation of the present methods to earlier procedures for finding transformations is discussed. The methods are illustrated with examples.