Hybrid choice model to disentangle the effect of awareness from attitudes: Application test of soft measures in medium size city

Abstract

The need to reduce private vehicle use has led to the development of soft measures aimed at re-educating car users through information processes that raise their awareness about the benefits of environmentally friendly modes, encouraging them to voluntarily change their travel choice behaviour (level of services characteristics being equal). It has been observed that these measures can produce enduring changes, being the result of mindful decisions. It is important then to try and understand what contributes to shape individuals? preferences in order to be able to define the best policy for fostering changes toward more pro-environmental modes.

Full text

Hybrid
choice
model
to
disentangle
the
effect
of
awareness
from
attitudes:
Application
test
of
soft
measures
in
medium
size
city
Eleonora
Sottile
a,
*,
Italo
Meloni
a
,
Elisabetta
Cherchi
b
a
University
of
Cagliari
—
CRiMM,
via
San
Giorgio
12,
Cagliari
09124,
Italy
b
Technical
University
of
Denmark
—
DTU,
Bygningstorvet
116B,
Lyngby
2800,
Denmark
A
R
T
I
C
L
E
I
N
F
O
Article
history:
Received
17
October
2015
Received
in
revised
form
15
July
2016
Accepted
14
September
2016
Available
online
23
September
2016
Keywords:
VTBC
Soft
measure
Stated
preference
Stress
CO
2
Hybrid
Choice
Models
A
B
S
T
R
A
C
T
The
need
to
reduce
private
vehicle
use
has
led
to
the
development
of
soft
measures
aimed
at
re-educating
car
users
through
information
processes
that
raise
their
awareness
about
the
benefits
of
environmentally
friendly
modes,
encouraging
them
to
voluntarily
change
their
travel
choice
behaviour
(level
of
services
characteristics
being
equal).
It
has
been
observed
that
these
measures
can
produce
enduring
changes,
being
the
result
of
mindful
decisions.
It
is
important
then
to
try
and
understand
what
contributes
to
shape
individuals’
preferences
in
order
to
be
able
to
define
the
best
policy
for
fostering
changes
toward
more
pro-environmental
modes.
The
objective
of
this
work
is
to
provide
empirical
evidence
of
the
effect
of
awareness
and
individual
attitudes
on
the
switch
from
car
driver
to
more
sustainable
modes
such
as
Park
and
Ride.
In
particular
we
attempt
to
discriminate
the
effect
of
awareness
due
to
the
information
provided
in
a
Stated
Preference
experiment
from
the
effect
of
individuals’
attitudes
toward
stress
and
social
norms
with
respect
to
sustainable
transport
modes.
The
case
study
refers
to
the
implementation
of
a
Voluntary
Travel
Behaviour
Change
programme
in
Cagliari
(Italy),
carried
out
with
the
purpose
of
promoting
the
use
of
the
light
rail
in
Park
and
Ride
mode.
To
account
for
all
these
effects
in
the
choice
between
car
and
Park
and
Ride
we
estimate
a
Hybrid
Choice
Model
where
the
discrete
choice
structure
allows
us
to
estimate
the
effect
of
awareness
of
environment
and
stress,
while
the
latent
structure
allows
us
to
estimate
the
effect
of
the
latent
effect
of
norms
and
attitudes
toward
environment
and
stress.
The
results
from
this
case
study
show
that
the
more
people
consider
the
information
about
stress
useful,
the
more
they
tend
to
behave
sustainably,
suggesting
the
importance
of
reporting
feedback
about
stress
in
the
personalised
travel
plan
to
promote
sustainable
mobility.
Interestingly,
the
information
about
pollution
has
instead
less
impact
in
shifting
behaviour
toward
sustainable
modes.
ã
2016
World
Conference
on
Transport
Research
Society.
Published
by
Elsevier
Ltd.
All
rights
reserved.
1.
Introduction
Road
traffic
is
now
the
main
culprit
of
air
pollution
in
urban
areas,
due
to
the
emissions
of
the
combustion
products
of
fuels
and
their
subsequent
chemical
transformation,
as
well
as
to
the
evaporation
of
unburned
hydrocarbons.
Transport
accounts
for
25%
of
CO
2
emitted
globally,
and
is
also
one
of
the
few
sectors
where
emissions
have
continued
to
grow,
oil
consumption
between
1973
and
2010
increasing
by
110 %
(Iea,
2011)
and
CO
2
by
44%.
(IEA,
2009;
Banister
et
al.,
2012).
Several
actions
and
measures
have
been
developed
to
try
to
mitigate
harmful
emissions.
These
mostly
refer
to
vehicles
technology
(greater
efficiency
both
in
terms
of
consumption
and
production
of
polluting
emissions),
type
of
fuel
(biofuel,
hydrogen,
and
electricity),
economic
tools
and
institutional
controls
(pricing
policies,
incentives,
taxes,
etc.)
information
and
communication
technologies
(ICT).
Although
powerful,
these
measures
have
not
been
proved
to
be
sufficient
to
solve
the
problem
(Schwanen
and
Lucas,
2011).
As
a
consequence,
in
recent
years,
research
has
increasingly
focused
the
attention
on
measures
and
policies
that
affect
individuals’
behaviour
and
in
particular
what
motivates
individu-
als’
decisions.
Providing
information
is
the
measure
most
used
to
promote
behaviour
change
(Abrahamse
and
Matthies,
2012).
As
reported
by
Ampt,
2003
“A
person
who
has
an
attitude
that
suggests
that
it
would
be
consistent
for
him
or
her
to
use
the
car
*
Corresponding
author.
E-mail
addresses:
elesottile@gmail.com
(E.
Sottile),
imeloni@unica.it
(I.
Meloni),
elich@transport.dtu.dk
(E.
Cherchi).
http://dx.doi.org/10.1016/j.cstp.2016.09.001
2213-624X/ã
2016
World
Conference
on
Transport
Research
Society.
Published
by
Elsevier
Ltd.
All
rights
reserved.
Case
Studies
on
Transport
Policy
5
(2017)
400–407
Contents
lists
available
at
ScienceDirect
Case
Studies
on
Transport
Policy
journal
homepage:
www.else
vie
r.com/locate
/cst
p

less
cannot
bring
about
behaviour
change
if
that
person
does
not
know
how
to
change”.
It
has
been
observed
that
measures
that
increase
individuals’
awareness
can
produce
enduring
changes,
being
the
result
of
mindful
decisions.
This
is
at
the
basis
of
the
concept
of
“Soft
Measures”,
also
referred
to
as
“Voluntary
Travel
Behaviour
Change”
(VTBC)
programmes
(Ampt,
2003)
or
“Smarter
Choices”
(Cairns
et
al.,
2004),
i.e.
programmes
aimed
at
motivating
the
voluntary
reduction
of
car
use.
VTBC
programmes
provide
information
typically
on:
a)
the
negative
(mainly
environmental)
effects
of
current
behaviour
and
b)
how
individuals
can
change
their
current
behaviour
to
mitigate
the
negative
effects.
As
opposed
to
supply
side
or
hard
measures,
where
policy
approaches
aimed
at
changing
mode
share
revolve
around
time
and
cost,
soft
measures
act
on
private
sphere
attributes.
In
particular,
not
only
on
how
a
person
perceives
the
choice
context
and
hence
its
structural
attributes
but
also
on
attitudes
and
on
the
propensity
to
behave
in
a
certain
way.
Indeed
soft
measures
rely
on
the
knowledge,
demonstrated
by
numerous
researchers,
that
an
objective
analysis
of
the
choice
context
and
its
characteristics
(time
and
cost)
does
not
always
suffice
to
evoke
change
in
travel
behaviour,
if
a
person
is
not
properly
informed
or
made
aware
of
the
most
beneficial
alternative
for
him-herself
or
for
society
as
a
whole.
These
measures
were
in
fact
devised
with
a
view
to
overcoming
some
of
the
shortcomings
of
the
traditional
approaches
to
understanding
travel
behaviour
and
behaviour
change.
Theoreti-
cally
speaking,
the
classical
engineering
approach
(structural/hard
measures)
to
travel
behaviour
research,
the
result
of
standard
microeconomic
theory
and
the
paradigm
of
rational
man,
has,
over
time,
been
combined
with
theoretical
models
from
social
psychology
and
behaviour
economics.
The
purpose
of
this
was
to
enhance
the
predictive
ability
of
economic
theory
by
providing
more
psychologically
plausible
foundations.
In
a
choice
context
where
time
and
cost
do
not
vary
(i.e.
no
changes
to
the
transport
system),
from
a
policy
standpoint
soft
measures
are
designed
to
raise
people’s
awareness
as
to
the
available
travel
alternatives
and
the
environmental
impact
of
travel
choice
and
as
a
result
to
change
attitudes
and
evoke
inclination
to
behave
differently
(Steg
and
Tertoolen,
1999).
The
information
provided
within
a
soft
measure,
as
well
as
the
quantitative
instrumental
feedback
(time
and
cost)
for
showing
the
personal
economic
benefits,
are
combined
with
feedback
about
the
societal
effects
concerning
the
environmental
and
health
spheres
(CO
2
and
calories
burned)
as
well
as
stress.
Briefly,
VTBC
programmes
aim
to
combine
the
traditional
behavioural
approach
with
cognitive
psychology
and
persuasion
principles.
The
objective
of
the
present
work
is
to
provide
evidences
about
the
effect
of
awareness
after
implementation
of
a
soft
measure
and
to
understand
the
relationship
between
awareness,
behaviour,
attitudes
and
norms
in
the
context
of
mode
choice.
The
research
provides
a
contribution
on
how
to
design
and
develop
a
VTBC
programme
to
promote
a
light
rail
service
and
on
how
to
study
the
extent
to
which
each
single
element
of
the
soft
measure
contributes
to
overall
awareness.
The
research
also
discusses
the
problem
of
measuring
the
effect
of
information
within
a
SP
experiments,
contributing
to
the
existing
literature
in
this
field
(see
a
recent
discussion
in
Cherchi
and
Hensher,
2015).
Finally,
results
allow
also
sketching
recommendation
for
policy
makers
on
the
information
and
feedback
that
are
more
effective
in
promoting
sustainable
Park
&
Ride
(P
&
R)
travel.
The
study
focuses
in
particular
on
the
effect
that
information
on
pollution
and
individual
stress
has
on
the
choice
to
shift
from
private
car
to
a
more
sustainable
P
&
R
mode
in
a
medium
sized
city
in
Italy.
The
promotion
campaign
was
based
on
the
knowledge
that
the
quality
of
the
environment
is
one
of
the
most
important
aspects
in
transport
problems
and
the
CO
2
emitted
is
probably
the
most
effective
(and
understandable)
measure
thereof.
However,
an
earlier
VTBC
programme
named
Casteddu
Mobility
Styles
(CMS),
the
first
to
be
implemented
in
Italy,
revealed
that
traffic
stress
was
an
interesting
variable
indicated
by
car
drivers.
Stress
has
a
very
negative
health
effect
on
modern
society,
and
the
stress
caused
by
traffic
conditions
is
very
subtle.
Traffic
stress
may
result
from
the
hassles
of
driving
and
parking,
the
potential
for
unintentional
injuries,
and
pecuniary
hardships
and
inconveniences
of
vehicle
maintenance
and
purchase
(Gee
and
Takeuchi,
2004).
Studies
of
commuting
have
identified
a
number
of
psychological
factors
that
govern
the
magnitude
of
stress
response,
including
control,
predictability,
time
urgency,
and
impedance
(Gottholmseder
et
al.,
2009;
Koslowsky,
1997).
High
levels
of
traffic
congestion
may
lead
to
elevated
physiological
stress
and
negative
effects
(Koslowsky
et
al.,
1995).
Notwithstanding
its
importance,
to
the
authors’
knowledge,
traffic
stress
has
never
been
considered
in
transport
studies.
The
ability
to
perceive,
or
to
be
conscious
of
something
and
to
react
to
it
(i.e.
awareness)
can
differ
from
one
person
to
another
depending
on
their
psychological
stance
toward
environment
and
stress.
Many
studies
have
accounted
for
the
effect
of
environmen-
tal
attitude
mainly
in
mode
choice
(Paulssen
et
al.,
2014)
or
in
the
choice
of
the
fuel-vehicle
(Daziano
and
Bolduc,
2013;
Glerum
et
al.,
2013;
Jensen
et
al.,
2013).
However,
other
latent
effects
other
than
attitude
are
relevant.
In
particular,
in
terms
of
environmental
awareness
and
the
information
provided,
personal
norms
measure
a
very
interesting
aspect
as
they
evaluate
the
moral
rule
(and
obligation)
that
lead
individuals
to
act
rightly
or
wrongly
towards
the
environment.
The
recent
literature
reports
a
couple
of
examples
of
the
effect
of
norms
in
a
choice
modelling
context
and
how
norms
stimulate
the
intention
to
use
public
transport
and
increase
actual
public
transport
usage
(Zhang
et
al.,
2015)
or
the
intention
to
switch
from
the
current
transport
mode
to
green
transport
modes
(Polydoropoulou
et
al.,
2015).
To
account
for
all
these
effects
in
the
choice
between
car
and
P
&
R
we
estimate
a
Hybrid
Choice
Model
(HCM)
where
the
utility
of
the
modes
depends
on
the
awareness
of
the
benefits
of
environmentally
friendly
travel
in
terms
of
emissions
reduction
and
stress
reduction.
In
particular
the
discrete
choice
structure
allows
us
to
estimate
the
effect
of
the
two
benefits
provided
on
the
mode
choice,
while
the
latent
structure
allows
us
to
estimate
the
effect
of
the
latent
attitudes
toward
environment
and
stress.
The
HCMs
include
three
latent
constructs:
attitude
toward
1)
stress
perceived
and
2)
information
received
about
stress
and
3)
personal
norms
with
respect
to
sustainable
transport
modes.
The
data
used
in
this
work
(gathered
during
a
campaign
to
promote
use
of
the
light
rail
in
Cagliari)
are:
1)
a
Stated
Preference
survey
where
the
information
provided
is
directly
included
as
attributes
in
the
SP
tasks
presented
to
the
individuals
and
2)
a
questionnaire
that
includes
a
set
of
questions
that
allow
us
to
evaluate
the
full
Theory
of
Planned
Behaviour
(TPB)
as
described
in
Ajzen
(1991)
to
measure
the
latent
aspects
through
items
that
allow
to
identify
the
indicators
of
the
measurement
equations
in
the
HCM.
2.
Case
study
The
context
chosen
for
the
experimental
analysis
is
the
corridor
that
connects
Cagliari
city
centre
with
the
city
of
Monserrato,
in
the
metropolitan
area
of
Cagliari,
where
in
2008
a
short
light
rail
line,
named
Metrocagliari,
went
into
operation.
The
Metrocagliari
is
6.3
km
long,
has
9
stops
and
190
trains
a
day
operate
in
each
direction
with
10-min
headway.
The
travel
time
between
the
two
terminals
is
18
min.
The
corridor
carries
150,000
round
car
trips/
day
and
around
56,000
inhabitants
live
within
500
m
from
the
9
stops.
Unfortunately
up
to
2010
only
5000
travellers/day
used
the
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407
401

light
rail,
about
75%
below
its
capacity.
Thus,
this
context
offered
a
good
opportunity
to
experiment
individualised
social
marketing
techniques
to
promote
the
use
of
an
existing
sustainable
mode.
The
survey
was
conducted
in
internet
and
announced
on
the
web
sites
of
the
municipalities
located
in
the
metropolitan
area
of
Cagliari.
The
methodology
followed
in
our
study
was
divided
into
three
steps:
1
A
typical
revealed
preference
(RP)
survey
was
carried
out
first
where
individuals
were
asked
to
describe
the
most
recent
trip
made
in
the
corridor
of
interest
for
the
study.
The
RP
data
were
mainly
gathered
to
customise
the
SP
cases,
as
it
is
particularly
important
to
adapt
the
alternatives
to
real
journey
experienced
by
each
respondent.
2
The
data
collected
with
the
RP
survey
were
carefully
analysed
to
identify
Prospective
Park
and
Riders
(PP
&
R),
i.e.
current
car
drivers
who
could
conveniently
use,
in
terms
of
travel
time
and
cost,
the
combination
of
car
and
Metrocagliari
(driving
by
car
from
the
origin
of
the
trip
to
the
light
rail
parking
area
and
taking
the
Metrocagliari
to
the
final
destination),
and
those
who
could
not
participate
due
to
any
kind
of
constraints.
3
Individuals
who
were
willing
to
participate
were
contacted
again
and
asked
to
fill
in
the
SP
survey
and
the
TPB
questionnaire.
In
total
1236
people
entered
the
web-page
and
started
the
questionnaire,
of
these
486
(41.97%)
completed
the
RP
question-
naire,
where
detailed
information
was
collected
about
the
last
trip
made
by
car
in
the
corridor
of
interest
(in-vehicle
travel
time,
walking
time
to
and
from
the
parking
space,
time
to
find
a
parking
space,
duration
of
parking
and
cost,
activities
performed
at
origin/
destination,
how
often
the
same
trip
is
made
a
month,
which
alternative
modes
were
not
available
for
that
specific
trip,
stops
inside
the
tour,
who
decides
to
make
that
trip
with
those
characteristics,
socio
economic
characteristics
of
the
individuals
and
their
family,
availability
to
participate
in
the
SP
experiment).
Unfortunately,
given
the
way
the
survey
was
announced,
we
were
not
able
to
send
reminders.
As
shown
in
Table
1,
among
those
who
filled
in
the
questionnaire,
the
target
sample,
i.e.
the
potential
park
and
riders,
were
only
197.
These
individuals
were
contacted
again
to
answer
the
second
part
of
the
survey
(SP
and
TPB).
The
response
rate
was
not
very
high
(32%),
though
many
reminders
were
sent
to
these
people.
The
final
sample
of
62
individuals
is
equally
distributed
between
males
and
females.
The
distribution
in
terms
of
age
sees
the
majority
(more
than
80%)
of
the
sample
between
31
and
60
years
old.
All
users
have
at
least
secondary
school
education
while
39%
had
a
Master
degree
and
18%
a
higher
degree.
Regarding
occupation,
in
accordance
with
age
and
education,
the
largest
part
(84%)
of
the
sample
is
occupied,
mostly
as
employees
(64.52%).
Unemployed
represent
less
than
10%,
while
the
percentage
of
students
is
even
lower,
due
to
the
fact
that
the
youngest
age
group
accounts
for
a
small
percentage
of
the
users.
Almost
half
the
respondents
(48.39%)
have
children.
The
average
number
of
household
members
is
2.79
with
2.1
cars
per
household
on
average.
2.1.
Stated
preference
experiment
The
use
of
RP
data,
which
are
normally
used
in
VTBC
programmes,
definitely
makes
it
more
difficult
to
determine
precisely
the
effect
of
soft
measures
and
discriminate
their
effect
from
others
that
may
be
present.
Contrariwise
SP
data
offer
the
advantage
of
being
able
to
determine
exactly
the
effect
of
the
measures
that
are
implemented.
Implementing
a
soft
measure
is
an
important
aspect
for
understanding
what
actually
contributes
to
shaping
people’s
preferences,
so
as
to
be
able
to
define
the
best
policy
for
evoking
change
towards
more
sustainable
travel
modes.
As
far
as
the
authors
are
aware
no
other
studies
are
reported
in
the
literature
that
implement
a
soft
measure
in
a
SP
survey.
In
our
study
we
focus
on
the
effect
of
soft
measures
on
the
change
of
mode
from
car
drivers
to
park
and
riders.
An
implicit
scenario
was
defined
for
the
status
quo,
where
(i)
parking
at
the
destinations
indicated
by
the
respondents
was
not
free
and
the
cost
increased
versus
the
current
situation
and
(ii)
congestion
on
the
roads
also
increased
causing
longer
in-vehicle
travel
time.
For
those
trips
where
the
final
destination
was
too
far
from
a
light
rail
stop,
the
hypothetical
scenario
also
included
an
extension
of
the
current
metro
line.
This
information
was
provided
in
the
introduction
to
the
SP
survey,
along
with
some
explanation
about
the
P
&
R
and
a
reminder
about
the
characteristics
of
travel
described
in
the
RP.
Since
the
SP
survey
is
a
mode
choice
experiment,
the
level-of-
service
(LOS)
attributes
relevant
are:
1
walking
time
(between
the
parking
space
and
the
metro
station
or
the
final
destination)
2
travel
time
by
car
and
by
light
rail
3
parking
time
4
headway
5
travel
costs
(fares
and
parking
costs
for
the
P&R
and
operational
and
parking
costs
for
the
car
alone)
6
transfers
Given
the
characteristics
of
the
light
rail
stations,
walking
time
from
the
car
park
to
the
metro
stops
is
on
average
3
min
with
a
very
small
variation
from
this
mean.
Thus
we
decided
not
to
vary
this
attribute
in
the
SP.
We
also
decided
to
keep
the
walking
time
in
the
car-only
alternative
equal
to
the
time
declared
in
the
RP
and
to
report
these
values
outside
the
SP
experiment.
None
of
the
individuals
in
the
sample
selected,
had
to
make
additional
transfers
other
than
that
from
the
car
to
the
light
rail
(i.e.
the
one
implicit
in
the
P
&
R),
therefore
a
transfer
attribute
was
not
included
in
the
SP.
Respondents
were
informed
in
the
introduction
that
the
alternative
P
&
R
implies
a
transfer.
All
the
other
attributes
(travel
time,
parking
time,
parking
cost
and
frequency)
were
included
in
the
SP
experiments,
with
3
levels
Table
1
Socio-economic
characteristics
in
the
sample.
Socio-economic
characteristics
N.
%
Gender
Female
30
48.39
Male
32
51.61
18–30
9
14.52
Age
31–40
23
37.10
41–60
27
43.55
61–80
3
4.84
Junior
School
Degree
0
0.00
High
School
Degree
27
43.55
Education
Master
Degree
24
38.71
Post
Lauream
11
17.74
Employee
40
64.52
Employment
status
Self-employed
12
19.35
Student
4
6.45
Unemployed
6
9.68
Child
30
48.39
Household
characteristics
No.
of
family
members
2.79
–
No.
of
family
cars
2.1
–
402
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407

each.
All
the
attributes,
except
frequency,
were
considered
generic
between
alternatives
and
values
pivoted
around
the
reference
values.
The
reference
values
were
slightly
varied
across
the
scenarios
so
as
not
to
bore
respondents.
We
pivoted
the
car
attributes
around
the
value
of
the
P
&
R
attribute
(i.e.
the
reference
value
was
in
all
cases
the
P
&
R).
As
mentioned,
the
RP
information
was
used
to
customise
the
SP
design.
Although
the
trips
selected
are
in
the
same
corridor,
they
can
have
quite
different
origins
and
destinations
and
they
are
made
at
different
times
of
the
day.
The
variation
in
travel
times
is
therefore
substantial.
We
then
analysed
the
sample
and
classified
respondents
into
groups
with
homogenous
LOS
characteristics
(travel
time
and
parking
cost
calculated
as
a
function
of
activity
duration)
and
identified
20
categories
(see
Table
2).
In
addition
to
the
3
LOS
attributes,
the
SP
also
included
2
information
attributes
“CO
2
reduction”
and
“stress
reduction”,
defined
with
2
levels
each.
An
orthogonal
design
was
used
and
the
final
design
with
27
choice
tasks
was
randomly
divided
into
three
blocks
of
9
choice
tasks
each.
The
use
of
information
attributes
in
the
SP
is
not
common
and
deserves
further
consideration.
The
major
challenges
in
including
the
information
about
CO
2
and
stress
as
attributes
concern
how
they
should
be
presented
to
respondents
in
order
to
be
clearly
understood.
It
is
widely
recognised
that
it
is
not
just
the
information
content
that
may
be
influential
but
travellers
may
also
be
affected
by
the
manner
in
which
information
is
presented
(Avineri
and
Waygood,
2013).
Mcfadden
(1997)
argues
that
the
majority
of
cognitive
anomalies
operate
through
errors
in
perception
that
arise
from
the
way
in
which
information
is
stored,
retrieved
and
processed.
We
devoted
special
attention
to
studying
the
best
way
to
present
the
soft
measures
in
the
SP
survey.
In
particular
we
tested
the
following
aspects:
1)
whether
to
use
images
alone,
only
text
or
both;
2)
the
type
of
information
that
should
be
included
in
the
text,
the
major
difficulty
lies
in
explaining
to
people
what
the
information
provided
means;
3)
the
type
of
context
to
be
included
in
the
images;
4)
whether
to
use
abstract
or
real
images
i.e.
cartoons
or
real
people.
We
focus
the
attention
on
how
this
information
should
be
presented
to
be
effective
and
we
carried
out
a
pilot
survey
to
test
the
following
aspects:

whether
to
use:

only
images

only
text

both

what
type
of
context
to
include
in
the
images:

general
context

specific
image
of
the
study
area

whether
to
use:

abstracted
(cartoon)

real
images

what
type
of
information
to
include
in
the
text:

percentage

absolute
value
We
defined
8
cases
based
on
the
combination
of
the
above
described
aspects,
with
and
without
text,
and
asked
a
sample
of
20
individuals
to
express
their
level
of
agreement/disagreement
(5-
point
Likert
scale)
for
each
case
by
means
of
the
four
adjectives:
1)
Clear;
2)
Relevant;
3)
Effective
and
4)
Original.
Finally
we
asked
the
respondents
to
choose
the
best
one.
We
found
that
the
best
way
to
present
information
was
a
combination
of
text,
percentage
and
realistic
images
specific
to
Cagliari:
-
18/20
individuals
chose
that
combination
-
Most
of
them
assigned
the
highest
score
in
the
Likert
scale
to
that
combination
for
all
four
adjectives.
2.2.
Attitudes
and
social
norm
questionnaire
According
to
the
Theory
of
Planned
Behaviour,
as
formulated
by
Ajzen
(1991),
attitudes,
subjective
norm
and
perceived
behavioural
control
influence
the
intention
to
perform
a
given
behaviour.
Attitudes
represent
the
degree
to
which
the
performance
of
the
behaviour
is
valued
while
subjective
norms
measure
the
perceived
social
pressure
to
engage
or
not
to
engage
in
the
behaviour,
while
perceived
behavioural
control
refers
to
the
perceptions
about
our
ability
to
perform
a
given
behaviour.
To
measure
all
the
items
included
in
the
TPB
we
defined
a
set
of
57
statements,
which
in
our
study
concerned:
public
transport
use,
car
use,
traffic
stress
and
environment.
A
factorial
analysis
was
conducted
in
order
to
identify
one
or
more
latent
dimensions
(called
factors
or
components)
underpinning
a
set
of
items
or
variables.
Two
types
of
factor
analysis
are
adopted
here:
Factor
Analysis
(FA)
and
Principal
Component
Analysis
(PCA),
that
differ
in
mathematical
terms.
The
PCA
analyses
all
the
variances
in
the
observed
variables
while
the
FA
only
the
shared
variance,
i.e.
it
does
not
consider
the
error
variance
in
the
specific
non-observed
variable
(Tabachnick
and
Fidell,
2007).
To
determine
the
factor-
ability
of
the
data
we
used
the
Kaiser-Meyer-Olkin
(KMO)
test.
To
determinate
the
number
of
factors
to
be
extracted
we
used
the
Kaiser
criterion
or
autovalue,
the
Scree
test
or
Cattell
test
(Cattell,
196 6)
and
Parallel
Analysis
(we
used
the
software
Monte
Carlo
PCA).
Once
the
latent
factors
had
been
identified,
we
obtained
a
measure
of
the
reliability
of
each
one
by
means
of
Cronbach’s
alpha
test.
This
coefficient
describes
the
internal
consistency
of
groups
of
items.
The
test
produces
a
value
of
between
0
and
1;
the
nearer
the
value
to
1
the
more
the
items
examined
behave
consistently
with
respect
to
each
item
of
every
factor.
The
factor
analysis
on
all
57
items
allowed
us
to
identify
16
latent
constructs.
For
this
paper
we
decided
to
focus
attention
on
the
three
issues
related
with
stress
and
environment,
thus
the
following
three
latent
variables
were
considered:
attitude
toward
1)
stress
perceived
and
2)
information
received
about
stress
and
3)
personal
norms
with
respect
to
sustainable
transport
modes.
Table
3
summarises
the
aspects
analysed
and
the
results
obtained.
Regarding
the
latent
variable
Info
Stress,
in
this
study,
we
take
into
account
just
one
factor
defined
by
indicator
1,
2,
3,
4,
6.
Table
2
LOS
characteristics.
Parking
Cost
[s/h]
Activities
duration
at
the
final
destination
Travel
Time
[min]
(0

10)
(11

15)
(16

25)
(26

35)
(>35)
Tot
0.5
s
d

1h
3.05%
5.08%
6.60%
1.52%
0.51%
16.75%
1.5
s
1
h
<d

2h
2.03%
5.58%
10.66%
6.60%
1.02%
25.89%
2.5

3
s
2
h
<d

4h
3.55%
2.54%
13.71%
6.09%
2.03%
27.92%
4
s
d
>
4h
3.05%
5.08%
14.21%
4.06%
3.05%
29.44%
Tot
11.68%
18.27%
45.18%
18.27%
6.60%
100.00%
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407
403

3.
Modelling
approach:
hybrid
choice
model
Following
Walker
(2001),
Ben
Akiva
et
al.
(2002)
and
Vij
and
Walker
(2016)
in
this
paper
we
use
a
Hybrid
Choice
Model
to
account
for
the
effect
of
the
awareness
about
the
benefits
of
environmentally
friendly
modes
and
the
effect
the
individual’s
attitude
toward
environment
on
mode
choice.
The
utility
of
the
modes
in
our
HCM
depends
on
the
awareness
of
the
benefits
of
environmentally
friendly
travel
in
terms
of
emissions
reduction
and
stress
reduction.
In
particular
the
discrete
choice
structure
allows
us
to
estimate
the
effect
of
the
two
benefits
provided
on
the
mode
choice,
while
the
latent
structure
allows
us
to
estimate
the
effect
of
the
latent
attitudes
toward
stress
and
the
social
norms
with
respect
to
sustainable
transport
modes.
Given
the
dimension
of
our
sample,
we
decided
to
include
one
latent
effect
at
a
time
because,
the
HCM
then
takes
the
structure
reported
in
Fig.
1.
Let
U
qj
be
the
utility
that
an
individual
q
associates
to
alternative
j.
As
in
the
typical
discrete
choice
models,
this
can
be
specified
as:
U
qj
¼
ASC
j
þ
u
j
LOS
qj
þ
b
j
SE
q
þ
t
j
I
qj
þ
l
j
ðaSE
q
0
þ
v
q
Þ
þ
e
qj
ð1Þ
Where
LOS
is
a
vector
of
level
of
service
characteristics
and
u
the
vector
of
coefficients
associated
to
them.
In
particular
in
this
work,
the
LOS
vector
includes:
-
travel
time
by
car
from
origin
to
destination,
-
travel
time
by
car
looking
for
parking
at
final
destination,
-
parking
cost,
-
travel
time
by
car
from
origin
to
park-and
ride.
-
is
the
travel
time
by
car
looking
for
parking
at
park-and
ride,
-
is
the
ticket
metro
cost,
-
is
the
waiting
time.
SE
is
a
vector
of
socioeconomic
characteristics
and
I
is
a
vector
of
information
provided,
while
b
and
t
are
the
respective
vectors
of
coefficients.
The
vector
of
socioeconomic
characteristics
includes
two
dummy
variables
for
age
between
18
and
30
and
age
between
31
and
40,
and
the
attributes
gender,
occupation
and
presence
of
child.
The
information
vector
instead
includes
two
dummy
variables:
one
the
information
provided
in
the
SP
about
emissions
reduction
and
one
for
the
information
provided
in
the
SP
about
traffic
stress
reduction.
ASC
is
a
vector
of
constants
specific
for
N-1
alternatives.
(a
SE
q
+
v
q
)
is
the
latent
variable
(LV)
that
depends
on
a
vector
of
SE’
(the
same
set
of
attributes
can
be
included
in
both
the
discrete
choice
and
the
latent
variable
model;
this
allows
to
test
that
the
latent
variable
is
not
acting
as
proxy
for
the
direct
effect
from
the
characteristics.),
and
a
is
the
associated
coefficient.
v
q
is
a
normal
distributed
error
term
with
zero
mean
and
standard
deviation
s
w
.
Table
3
Latent
variables
Items.
LV
Questions
Items
(Indicators
of
LV)
KMO
No.
Factors
Variance
explained
Alpha
factor
1
Attitude
toward
Stress
Perceived
Which
of
the
following
aspects
stresses
you
more
1.
Traffic
2.
Car
maintenance
3.
Accidents
4.
Difficulties
in
finding
a
parking
place
5.
Traffic
noise
0.747
1
46%
0.703
Attifude
toward
Info
received
about
Stress
Do
you
think
that
receiving
information
about
the
level
of
stress
associated
with
driving
can:
1.
Be
important
but
not
as
much
as
travel
costs
and
times
2.
Increase
people’s
awareness
as
to
the
negative
effects
associated
with
car
use
3.
Provide
an
incentive
to
reduce
car
use
4.
Make
car
users
reflect
about
the
possibility
of
switching
to
public
transport
5.
Has
absolutely
no
influence
on
the
choice
of
travel
mode
6.
Be
considered
useless
7.
Make
car
users
switch
to
public
transport
0.736
2
67%
0.820
Personal
Norm
with
respect
to
susttainable
modes
Do
you
agree
or
disagree
with
the
following
statements
1.
Regardless
of
what
other
people
do,
I
feel
a
moral
duty
to
travel
in
an
environmentally
more
sustainable
way
2.
Regardless
of
what
other
people
do,
I
feel
bad
if
I
am
unable
to
travel
in
an
environmentally
more
sustainable
way
3.
Regardless
of
what
other
people
do,
I
feel
good
if
I
do
not
use
the
car
a
lot
0.548
1
57%
0.588
Fig.
1.
Behavioural
Framework
for
the
choice
model
with
the
latent
variable
“stress
perceived”.
404
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407

Table
4
Model
results.
Variables
HCM
1
(Info
Stress)
HCM
2
(Personal
Norm)
HCM
3
(Stress)
Estimate
Robust
t-test
Estimate
Robust
t-test
Estimate
Robust
t-test
Discrete
choice
model
Constant_P
&
R
18.900
4.97
3.330
2.81
6.75
5.04
P
&
R
and
Car
LOS
Attributes
Travel
time
0.109
4.50
0.068
3.53
0.08
4.04
Time
looking
for
parking
0.087
1.58
0.046
1.0 6
0.04
0.93
Parking
cost
0.711
5.90
0.498
5.15
0.51
4.81
Travel
time
*
Time
search
for
parking
0.018
1.14
0.024
1.62
0.047
3.11
P
&
R
LOS
Attribute
Waiting
time
0.162
2.88
0.1
2.16
0.13
2.48
P
&
R
Information
CO
2
Reduction
Stress
Reduction
0.535
0.699
1.83
2.49
0.345
0.489
1.46
2.14
0.371
0.546
1.48
2.20
P
&
R
Individual
characteristics
Age
18–30
2.010
5.62
Age
31–40
1.210
4.88
1.24
4.26
Female
0.719
2.63
0.836
3.75
0.778
3.31
Self-employed
1.65
4.36
Presence
of
children
0.77
3.35
1.13
3.79
P
&
R
Latent
Variable
Constant_LV
3.51
47.44
3.14
63.05
3.93
18.82
Info
Stress
5.55
5.38
Personal
Norm
0.299
1.6 6
Stress
1.11
5.18
Latent
variable
model
LV-Individual
characteristics
Age
31–40
0.069
1.79
0.38
2.16
Age
>
41
0.333
4.21
0.052
6.95
0.68
5.29
Female
0.093
3.25
0.205
2.09
Self-employed
0.144
3.83
0.623
4.90
Number
of
household
members
0.16
3.57
Presence
of
children
0.145
3.00
Number
of
car/household
0.063
2.91
0.211
15.96
Random
term
1.23
7.64
0.515
8.28
0.22
5.03
Info
stress
Ind2:
Constant
3.27
3.87
Coeff_LV
1.98
8.05
Sigma
0.143
3.12
Ind3:
Constant
8.47
4.33
Coeff_LV
3.45
6.13
Sigma
0.496
9.06
Ind4:
Constant
7.89
4.13
Coeff_LV
3.36
6.06
Sigma
0.795
10.25
Ind6:
Constant
5.87
4.06
Coeff_LV
2.58
6.20
Sigma
0.511
15.67
Personal
Norm
Ind2:
Constant
4.93
10.43
Coeff_LV
2.07
16.14
Sigma
2.67
67.67
Ind3:
Constant
0.646
2.43
Coeff_LV
0.776
10.80
Sigma
0.033
1.21
Stress
Perceived
Ind2:
Constant
0.21
0.73
Coeff_LV
0.82
9.01
Sigma
0.141
3.17
Ind3:
Constant
0.07
0.21
Coeff_LV
0.61
6.03
Sigma
0.094
2.30
Ind4:
Constant
0.526
1.80
Coeff_LV
0.861
10.22
Sigma
0.132
2.74
Ind5:
Constant
0.57
2.27
Coeff_LV
0.864
10.70
Sigma
0.01
0.22
L(max)
3459.541
2455.202
4334.793
Sample
size
513
513
513
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407
405

l
is
the
coefficient
associated
to
LV.
Finally,
e
qj
is
the
error
term
identically
and
independently
distributed
extreme
value
type
1.
The
statements
measured
through
the
TPB
questionnaire
are
used
as
indicator
of
the
latent
variables
and
are
linked
to
it
with
the
following
measurement
equation:
I
qk
¼
g
k
þ
z
k
LV
q
þ
u
qk
k
¼
1;
:::;
K
ð2Þ
Where
I
qk
is
the
k-th
indicator
for
the
latent
variable,
g
k
is
the
intersect,
z
k
is
the
coefficient
associated
to
the
latent
variable
(g
and
z
are
normalized
to
zero
and
one
for
the
first
indicator
for
identification
purposes),
and
y
qk
is
the
normal
distributed
error
term
with
zero
mean
and
standard
deviation
s
y
.
The
structural
equation
for
the
discrete
choice
is
defined
by
the
dummy
variable,
d
wjnt
,
that
takes
the
value
one
if,
the
alternative
j
has
the
highest
utility
among
all
the
alternatives
in
the
choice
set
C
q
of
individual
q:
d
w
jnt
¼0
otherwise
1
ifU
w
jnt
>U
w
;8i6¼j;ði;jÞ2C
q
(ð3Þ
While
the
distributions
of
the
latent
variable
and
the
indicator
are
respectively:
f
LV
ðLV
q
jSE
q
;
a;
s
v
Þ
¼1
s
v
fLV
q

aSE
q
s
v


f
I
ðI
q
jLV
q
;
g;
z;
s
n
Þ
¼1
s
nk
fI
q
k

g

zLV
q
s
nk

ð4Þ
The
models
can
be
estimated
by
maximum
likelihood
estima-
tion.
The
choice
probabilities
are
given
by:
P
qj
¼Z
v
P
qj
ðLV
q
ðv
q
ÞÞf
LV
ðv
q
Þf
I
ðLV
q
ðv
q
ÞÞf
ðvÞdv
ð5Þ
Models
are
estimated
using
PythonBiogeme
(Bierlaire
and
Fetiaris,
2009).
4.
Results
Table
4
gives
the
results
of
the
three
HCM
estimated
including
one
latent
effect
at
a
time.
These
initial
results
are
very
interesting.
Firstly
we
note
that
the
signs
of
all
the
parameters
of
the
LOS
attributes
are
in
agreement
with
microeconomic
theory
and
are
all
different
from
zero,
with
a
significance
level
of
95%
except
for
the
time
spent
looking
for
parking
maybe
because
it
has
a
very
low
value
for
the
P
&
R
alternative.
The
information,
describing
the
effect
of
the
soft
measure
implemented
through
the
SP,
has
been
included
as
specific
to
the
P
&
R
alternative.
We
found
that
the
parameters
associated
thereto
are
positive,
indicating
that
the
utility
of
P
&
R
increases
with
the
level
of
awareness
achieved
through
the
information
received
about
the
P
&
R
alternative.
In
all
models
the
coefficient
of
stress
is
significant
at
95%,
while
for
the
coefficient
of
CO
2
we
can
reject
the
hypothesis
that
it
is
different
from
zero
only
at
93%,
86%
and
84%
(in
a
one-tail
test)
respectively
in
the
3
models
presented.
All
three
latent
variables
tested
affect
the
choice
of
P
&
R,
and
in
particular
the
LVs
related
to
the
stress
are
highly
significant.
Further,
as
can
be
seen,
the
effect
of
the
LVs
is
in
line
with
the
effect
of
the
two
information
variables
measured
inside
the
SP
experiment.
Particularly,
the
more
individuals
consider
receiving
information
about
stress
useful,
the
more
they
tend
to
behave
sustainably
choosing
P
&
R
(positive
sign
of
the
latent
variables).
Lastly,
those
aspects
associated
with
stress
would
appear
to
have
a
greater
influence
on
travel
choice
than
environmental
aspects
(the
information
received
about
the
reduction
of
CO
2
emissions
is
not
as
significant
in
mode
choice
as
the
information
on
stress
and
the
same
can
be
said
for
personal
norms
compared
to
the
other
two
variables).
Indeed
individuals
are
reasonably
expected
to
be
more
sensitive
to
effects
that
have
immediate
repercussions
such
as
stress,
rather
than
to
long
term
effects
such
as
environmental
pollution.
We
tested
many
systematic
effects
other
than
the
ones
included
in
the
results,
but
we
only
left
those
significant
at
0.01%,
in
order
not
to
over
fit
the
models.
We
have
tried
also
to
estimate
the
three
latent
variables
at
the
same
time
in
one
model
but
the
model
doesn't
converge.
Interestingly
the
effectiveness
of
the
information
provided
can
easily
be
deduced
simply
by
analysing
the
number
of
times
P
&
R
has
been
preferred
over
the
car
in
those
cases
where
information
variables
are
included.
As
can
be
seen
in
the
following
graph,
the
highest
percentage
of
P
&
R
choices
coincides
with
the
provision
of
both
kinds
of
information,
followed
by
information
on
stress
alone,
information
on
CO
2
alone
and
lastly
no
information
at
all
(Fig.
2).
5.
Conclusions
In
the
present
work,
conducted
within
the
framework
of
voluntary
travel
behaviour
change
programmes,
the
effect
of
implementing
soft
travel
demand
management
measures
on
mode
choice
has
been
studied
through
the
estimation
of
hybrid
models.
In
particular
it
has
been
possible
to
discriminate
the
effect
of
these
measures
from
the
effect
that
psychological
aspects
can
have
on
choice
behaviour.
The
study
confirmed
the
importance
of
distinguish
the
effect
of
information
provision
and
the
effect
of
the
latent
aspects
underpinning
behaviour
on
opting
for
P
&
R
rather
instead
of
the
private
car.
68.55
66.13
72.58
58.06
28.23
29.84
24.19
39.11
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Stress
(124 tota
l n. obs)
CO2
(124 tota
l n. obs)
Both
(62
total
n.
obs
)
Neith
er
(28
4 tota
l n. obs)
Choice percentage
Cases_Information
Choice on
function
of
cases
with/
without
information
provided
Neith
e
r
Car
P&R
Fig.
2.
Choice
versus
information
provided.
406
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407

The
paper
describes
a
methodology
for
designing,
presenting
and
implementing
information
measures
and
presents
a
discus-
sion
of
those
aspects
that
can
enhance
or
undermine
the
SP
efficacy.
In
particular,
the
analysis
of
the
type
of
attributes
of
the
alternatives
and
how
they
needs
to
be
presented
proved
to
be
especially
useful
for
designing
the
personalised
travel
plans
and
for
feedback
perception.
The
results
obtained
with
the
models
estimated
confirm
that
individuals
indeed
perceive
information
as
beneficial.
A
positive
effect
on
the
utility
of
choosing
P
&
R
was
observed
for
all
the
information
tested.
Omitting
these
factors
from
VBTC
programmes
could
actually
undermine
the
effectiveness
of
the
promotional
campaign.
Further,
we
found
that
treating
the
psychological
aspects
as
latent
variables,
suitably
measured
using
indicators,
significantly
affect
mode
choice.
More
interestingly,
we
also
found
that
the
direct
information
has
a
strong
effect
beside
the
individuals’
attitude
and
social
norms.
The
model
results
confirmed
the
importance
of
simultaneously
investigating
the
effects
of
the
psychological
aspects
underlying
behaviour.
Particularly,
the
more
people
consider
the
information
about
stress
useful,
the
more
they
tend
to
behave
sustainably,
choosing
P
&
R.
The
stress-related
aspects
appear
to
have
a
greater
influence
than
the
environmental
ones,
while
the
information
about
CO
2
emissions
is
not
as
significant
in
mode
choice
as
information
about
stress.
This
result
suggests
it
is
important
to
indicate
feedback
about
stress
in
the
personalised
travel
plan
and
feedback
to
be
presented
to
car
drivers,
though
this
may
jeopardise
efficacy.
This
paper
intends
to
highlight
the
importance
of
evaluating
the
behaviour
process
in
a
comprehensive
manner
to
enhance
the
efficacy
of
voluntary
travel
behaviour
change
programmes
and
for
designing
personalised
travel
plans.
The
incorrect
evaluation
of
both
the
procedures
for
designing
and
implementing
soft
measures
and
of
all
those
attributes
that
affect
travel
behaviour
could
undermine
the
effectiveness
of
the
measures
and
from
a
modelling
standpoint,
lead
to
incorrect
demand
forecasting.
We
are
aware
of
the
limitation
of
this
study
due
to
the
small
number
of
participants,
but
it
is
difficult
to
get
big
samples
in
this
type
of
complex
studies,
and
this
type
of
studies
are
indeed
scares.
Anyway
we
think
that
our
study
represents
an
important
contribution
to
the
field
and
an
interesting
case
study
for
the
future
work.
Acknowledgement
The
Authors
are
grateful
to
the
Sardinian
Government
for
funding
the
project
(Legge7/2007).
References
Abrahamse,
W.,
Matthies,
E.,
2012.
Informational
strategies
to
promote
pro-
environmental
behaviours:
changing
knowledge,
awareness,
and
attitudes.
Environ.
Psychol.
Introd.
223–232.
Ajzen,
I.,
1991.
The
theory
of
planned
behaviour.
Org.
Behav.
Human
Decis.
Process.
50
(2),
179 –211.
Ampt,
E.,
2003.
Voluntary
household
travel
behaviour
change
-
theory
and
practice.
10th
International
Association
of
Travel
Behaviour
Research
Conference,
August,
2003,
Lucerne,
Switzerland,
pp.
10–15.
Avineri,
E.,
Waygood,
E.O.D.,
2013.
Applying
valence
framing
to
enhance
the
effect
of
information
on
transport
related
carbon
dioxide
emissions.
Transp.
Res.
Part
A
48,
31–48.
Banister,
D.,
Schwanen,
T.,
Anable,
J.,
2012.
Introduction
to
the
special
section
on
theoretical
perspectives
on
climate
change
mitigation
in
transport.
J.
Transp.
Geogr.
24,
467–470.
Ben
Akiva,
M.,
Mcfadden,
D.,
Train,
K.,
Walker,
J.L.,
Bhat,
C.R.,
Bierlaire,
M.,
Bolduc,
D.,
Boersch-Supan,
A.,
Brownstone,
D.,
Bunch,
D.,
Daly,
A.,
DE
Palma,
A.,
Gopinath,
D.,
Karlstrom,
A.,
Munizaga,
M.,
2002.
Hybrid
choice
models:
progress
and
challenges.
Market.
Lett.
13
(3),
163–17 5.
Bierlaire,
M.,
Fetiarison,
M.,
2009.
Estimation
of
discrete
choice
models:
extending
BIOGEME.
Presented
at
9th
Swiss
Transport
Research
Conference,
Monte
Veritá,
Switzerland.
Cairns,
S.,
Sloman,
L.,
Newson,
C.,
Anable,
J.,
Kirkbride,
A.,
Goodwin,
P.,
2004.
Smarter
choice,
changing
the
way
we
travel.
Final
Report
of
the
Research
Project:
‘The
Influence
of
Soft
Factor
Interventions
on
Travel
Demand’.
Published
by
the
Department
for
Transport,
London
on
the
‘Sustainable
Travel’
(section
of
www.
dft.gov.uk).
Cattell,
R.B.,
196 6.
The
screen
test
for
the
number
of
factors.
Multivar.
Behav.
Res.
1,
245–276.
Cherchi,
E.,
Hensher,
D.A.,
2015.
Stated
preference
surveys
and
experimental
design:
an
audit
of
the
journey
so
far
and
future
research
perspectives.
Transp.
Res.
Procedia
11,
154 –164.
Daziano,
R.A.,
Bolduc,
D.,
2013.
Incorporating
pro-environmental
preferences
towards
green
automobile
technologies
through
a
Bayesian
hybrid
choice
model.
Transportmetrica
Part
A
9
(1),
74–10 6.
Gee,
G.C.,
Takeuchi,
D.T.,
2004.
Traffic
stress,
vehicular
burden
and
well
being:
a
multilevel
analysis.
Soc.
Sci.
Med.
59,
405–414.
Glerum,
A.,
Stankovikj,
L.,
Thémans,
M.,
Bierlaire,
M.,
2013.
Forecasting
the
demand
for
electric
vehicles:
accounting
for
attitudes
and
perceptions.
Transp.
Sci.
48
(4),
483–499.
Gottholmseder,
G.,
Nowotny,
K.,
Pruckner,
G.J.,
Theurl,
E.,
2009.
Stress
perception
and
commuting.
Health
Econ.
18,
559–576.
IEA,
2009.
How
the
Energy
Sector
can
Deliver
on
a
Climate
Agreement
in
Copenhagen:
Special
Excerpt
of
the
World
Energy
Outlook
for
the
Bangkok
UNFCCC
Conference.
International
Energy
Agency,
Paris.
Iea,
2011.
Key
World
Energy
Statistics.
International
Energy
Agency,
Paris.
Jensen,
A.,
Cherchi,
E.,
Mabit,
S.,
2013.
On
the
stability
of
preferences
and
attitudes
before
and
after
experiencing
an
electric
vehicle.
Transp.
Res.
Part
D:
Transp.
Environ.
25,
24–32.
Koslowsky,
M.,
Kluger,
A.N.,
Reich,
M.,
1995.
Commuting
Stress:
Causes,
Effects,
And.
Methods
of
Coping.
Plenium,
New
York.
Koslowsky,
M.,
1997.
Commuting
stress:
problems
of
definition
and
variable
identification.
Appl.
Psychol.
46,
153–173.
Mcfadden,
D.L.,
1997.
Rationality
for
Economists
Presented
at
the
NSF
Symposium
on
Eliciting
Preferences.
Berkeley,
California.
Paulssen,
M.,
Temme,
D.,
Vij,
A.,
Walker,
J.L.,
2014.
Values,
attitudes
and
travel
behavior:
a
hierarchical
latent
variable
mixed
logit
model
of
travel
mode
choice.
Transportation
41,
873–888.
Polydoropoulou,
A.,
Kamargianni,
M.,
Tsirimpa,
A.,
Krousouloudi,
N.,
2015.
A
hybrid
Choice
Model
to
capture
the
effects
of
subjective
norms
and
attitudes
on
travel
mode
switching
behavior.
IATBR
2015.
Windsor.
Schwanen,
T.,
Lucas,
K.,
2011.
Understanding
auto
motives.
In:
Lucas,
K.,
Blumenberg,
E.,
Weinberger,
R.
(Eds.),
Auto
Motives:
Understanding
Car
Use
Behaviours.
Emerald,
Bradford.
Steg,
L.,
Tertoolen,
G.,
1999.
Sustainable
transport
policy:
the
contribution
from
behavioural
scientists.
Public
Money
Manag.
19
(1),
63–69.
Tabachnick,
G.G.,
Fidell,
L.S.,
2007.
Experimental
Designs
Using
ANOVA.
Duxbury,
Belmont,
CA.
Vij,
A.,
Walker,
J.L.,
2016.
How,
when
and
why
integrated
choice
and
latent
variable
models
are
latently
useful.
Transp.
Res.
Part
B:
Methodol.
90,
192 –217.
Walker,
J.L.,
2001.
Extended
Discrete
Choice
Models:
Integrated
Framework,
Flexible
Error
Structures,
and
Latent
Variables.
PhD.
Department
of
Civil
and
Environmental
Engineering,
Massachusetts
Institute
of
Technology.
Zhang,
D.,
Schmöcker,
J.D.,
Fujii,
S.,
2015.
Social
norms
and
public
transport
usage:
empirical
study
from
Shanghai.
Transportation
doi:http://dx.doi.org/10.1007/
s11116-015-9625-y.
E.
Sottile
et
al.
/
Case
Studies
on
Transport
Policy
5
(2017)
400–407
407