Stated Preference Surveyfor New Smart Transport Modesandservicesdesignpilotstudyandnewrevision

Transportation Systems
Working Paper Series
Stated Preference Survey for New Smart

Transport Modes and Services:
Design, Pilot Study and New Revision
Paper# ITS-SCUSSE-09-02
March 2009
Lang Yang, Charisma F Choudhury & Moshe Ben-Akiva

Massachusetts Institute of Technology
João Abreu e Silva & Diana Carvalho

Instituto Superior Tecnico
Stated Preference Survey for New Smart Transport Modes and
Services: Design, Pilot Study and New Revision
MIT Portugal Program

Transportation Systems Focus Area
Research Domain:
Intelligent Transportation Systems
Research Project:
Smart Combination of Passenger Transport Modes and Services in Urban Areas for
Maximum System Sustainability and Efficiency (SCUSSE)
Paper#:
ITS-SCUSSE-09-02
March 2009
Lang Yang ([email protected]), Charisma F Choudhury ([email protected]),

Moshe Ben-Akiva ([email protected])
Intelligent Transportation Systems Lab

Department of Civil and Environmental Engineering
Massachusetts Institute of Technology
João Abreu e Silva ([email protected]), Diana Carvalho ([email protected])
Instituto Superior Tecnico

Universidade Tecnica de Lisbon, Protugal
This publication was made possible by the generous support of the Government of
Portugal through the Portuguese Foundation for International Cooperation in Science,
Technology and Higher Education and was undertaken in the MIT-Portugal Program.
March 2009 1
Contents
1 Introduction............................................................................................................... 3
2 Preparatory Phase and Pilot Study ............................................................................. 6
2.1 Development of the SP Survey ........................................................................... 6
2.1.1 Focus Group Discussion .............................................................................. 6
2.1.2 Framing the Questions of the SP Survey ...................................................... 7
2.1.3 Experimental Design and Test.................................................................... 12
2.2 Data.................................................................................................................. 12
2.2.1 General Findings........................................................................................ 13
2.2.2 Respondent Characteristics ........................................................................ 14
2.2.3 RP Mode Choice Data................................................................................ 15
2.2.4 SP Mode Choice Data ................................................................................ 17
2.2.5 Departure Time Choice Data...................................................................... 18
2.3 Model Estimation ............................................................................................. 19
2.3.1 Mode Choice Model .................................................................................. 20
2.3.2 Departure Time Choice Model ................................................................... 27
2.4 Attitudes to Car and Public Transportation........................................................ 30
2.5 Information Services......................................................................................... 35
3 Revised Stated Preference Survey............................................................................ 36
3.1 Main Adjustments from Pilot Study.................................................................. 36
3.2 Updated SP Scenarios with Multidimensional Choice....................................... 36
3.3 Updates in the Experimental Design ................................................................. 41
4 Future Plans and Challenges.................................................................................... 43
References .................................................................................................................... 45
Appendix A................................................................................................................... 47
March 2009 2
1 Introduction
New Smart Transport Modes and Services
The objective of SCUSSE (Smart Combination of Passenger Transport Modes and

Services in Urban Areas for Maximum System Sustainability and Efficiency) is to
conceive, organize and simulate the implementation of new smart transport modes and
services to optimize integration with lifestyles and to improve the sustainability and
efficiency for urban transport systems, including the institutional design required for
and/or enabled by the deployment of innovative services.
The existing modes of day-to-day travel in Lisbon primarily include car (as a driver,
normal passenger or car-pool user), bus and heavy modes like subway, train and ferry.
Some of the journeys are made by a combination of these modes as well (e.g. bus and
heavy mode, car and heavy mode) etc. SCUSSE investigates different aspects of
improving the existing modes as well as introduction of new smart transport modes in
order to improve the efficiency of urban transport systems. The new transport modes
investigated in this study include the following:
! One-way car rental

This service provides access to light electric vehicles available at nearby parking lots
throughout the city placed in closely spaced intervals. Travelers can simply walk to a
nearest lot, swipe a card to pick up a vehicle, drive it to the lot nearest to your destination,
and drop it off there. The insurance, service, repair, fuel and parking costs are included in
the rental price and service package. Parking spots are also guaranteed for the service
users at the destinations. This mode is more convenient and less expensive compared to a
typical rental car. Service users need to keep some money (or provide credit card
number) as a caution.
! Shared taxi
This service provides a taxi service with call-centre dispatch access. Upon boarding, a
passenger is asked whether he/she is willing to share the taxi with other passengers who
have similar routes. If he/she agrees, other passengers will board the taxi until the vehicle
capacity is reached. The fare will be determined based on the most convenient distance
(on a solo trip) and the time penalty endured for the sake of the other passengers. This
mode is less expensive compared to a typical taxi service.
! Express minibus
This service provides a minibus service with fixed routes, few stops near the origin and
the destination (2 or 3 at most), and a significant stretch in between. The minibus has a
regular and pre-programmed schedule. This transport mode mainly focuses on frequent
commuters that live and work close to rather convenient places, who can share a
collective pick-up location and destination. There may be a few places available for
occasional riders as well. The minibus service is only available during peak periods 8:00
to 10:30 in the morning and 16:30 to 20:00 in the afternoon.
! Park and ride with child drop-off
March 2009 3
This service provides access to park and ride facilities with reserved parking spaces,
where commuters can leave their car and board a subway/train/ferry. In addition, there is
panoply of other services such as children drop-off (where children less than 10 years old
can be dropped off to be picked up by professional tutors). The tutors will be reliable
persons chosen collectively (either school teachers or professional people) and will take
care of the children before taking them to their school in school buses. There will be a
monthly charge associated with the service.
New traffic management methods include the following:

! Congestion pricing policies that apply charges based on departure time
! Adjustment of parking pricing and enforcement.
! New information services, including Internet or cell phone based traffic
information, and GPS navigation systems that can be used to predicate travel time
and reduce the time variability.
Motivation of Stated Preference Survey
Stated Preference (SP) surveys, also called self-stated preferences for market products or
services, have been widely applied in the areas of marketing and travel demand
modeling, separately or jointly with Revealed Preference (RP) surveys with observed
choices of product purchase or service use. It is an efficient method to analyze
consumers’ evaluation of multi-attributed products and services, especially when there
are hypothetical choice alternatives and new attributes.
In the case of Lisbon, Portugal, there are no Revealed Preference (RP) data for the
proposed innovative transport modes and information services, and there are no existing
congestion pricing strategies in urban areas. Therefore, a Stated Preference (SP) survey
must be well designed and implemented for our objectives, such as
! To evaluate acceptability of innovative modes and services

! To quantify sensitivity to level of service by varying values of access time,
waiting time, travel time, cost etc.
! To measure willingness-to-pay
! To investigate effects of attitudes and perceptions.
Challenges and Complexity
There are current transport modes, innovative transport modes, and also new
combinations of existing and innovative. The result is a large choice set for a particular
origin-destination. The ten most probable choice sets are as follows1:
! Driving alone in a private car
! Carpool
! Bus
! Heavy mode
! Bus and heavy mode
1
The less probable combinations are excluded for simplicity
March 2009 4
! One-way car rental
! Shared taxi
! Express minibus
! Park & ride (with or without child drop-off), and
! One-way car rental with heavy mode.
There are numerous attributes that need to be considered for these alternatives.
Furthermore, new smart services introduced in Lisbon include congestion pricing,

adjustment of parking pricing and enforcement, and information services such as Internet,
cell phone and GPS navigation systems. They will affect the accuracy of travel time
predication and people’s time-of-day preference. The SP survey should include
corresponding attributes and alternatives to capture such effects.
In addition, these alternatives are not uniform in format (e.g. waiting time is applicable to
public transport modes only, parking is associated with car only etc.). Therefore, the
organization and presentation of these alternatives and attributes is a challenge for the
Stated Preference (SP) survey.
An additional challenge was to accommodate the availability of modes (not all modes are
available for all origin destination pairs) and the base value of attributes for a certain
mode (which are context dependent). Also, the success of the survey depended on how
realistically the survey can be presented so that potential biases in the data are minimized.
The process of developing the SP survey is summarized in seven steps:

1. Defining important attributes
2. Designing the questionnaire of SP survey
3. Experimental design
4. Testing with synthetic data
5. Pilot study and analysis
6. Revising SP survey
7. Implementation of Internet survey and supplemental presential survey
Structure of This Paper
The remainder of this paper is organized as follows.
Section 2 describes the preparatory and testing phase of the main survey: this includes
summary of the focus group study and its findings, detailing the experimental design
method, developing the Pilot version of the Stated Preference survey and testing it using
a small number of respondents. Findings of simple discrete choice models estimated with
Pilot data are also presented in this context.
Section 3 discusses the revised version of the SP survey and associated design
modifications.
Section 4 talks about future plans and modeling challenges.
March 2009 5
2 Preparatory Phase and Pilot Study
The SP survey included ten alternatives, five of which are new smart alternatives or
combination of exiting and smart alternatives. Each of the alternatives involves numerous
attributes and a focus group study was conducted first to identify the more preferred
attributes.
The findings of this focus group study were the basis of designing the preliminary or
Pilot version of the survey, which was tested with 150 respondents.
This section presents summary of the focus group study and its findings, the framework
of the survey and the selected experimental design method and statistical analysis of the
Pilot data followed by findings of simple discrete choice models estimated with this data.
2.1 Development of the SP Survey
2.1.1 Focus Group Discussion
A focus group discussion was conducted March 2008 in Lisbon, Portugal. The objectives
were to find aspects of public transport, car, the new alternative modes and services that
could act as attraction or repulsion factors, to identify important attributes characterizing
the new services that may be used in the SP survey, and to identify potential attitudinal
aspects that could be included in the SP survey.
The main findings of the focus group discussion are as follows (see Viegas et al, 2008 for
a more in-depth description of this focus group results).
Carpool - not culturally adapted to the Portuguese
! Advantages: low costs and environmentally friendly
! Disadvantages: loss of independence and the possibility of conflicts
Shared Taxis - good receptivity
! Advantages: low price, environmentally friendly, good option when public
transport was not frequent
! Disadvantages: long travel time, lack of reliability and security
Minibus - good receptivity
! Advantages: comfort
! Disadvantages: less flexibility and high costs
Park & Ride with child drop-off - skepticism
! Advantages: connected with park and ride, good option for people who did not
mind leaving their children with others
! Disadvantages: lack of security for children, lack of confidence in tutors and
drivers
Congestion charge - mixed feelings
! In principle people agreed with this measure
! Approval depended on how to use the collected money and the necessity to
provide some kind of support to the ones that have an absolute need to use their
cars
Information systems
March 2009 6
! Reliability and precision was an important issue
! Information should provide data for inter-modal options
When choosing among transport modes and services, we found the important attributes
for the local residents were travel time, time variability, travel cost, frequency, the
reliability of tutors for park & ride with child drop-off, etc. There also existed some
attitudinal factors possibly to affect people’ preference, such as comfort, privacy,
flexibility, convenience, environmental friendly, and security.
2.1.2 Framing the Questions of the SP Survey
The large questionnaire of SP survey was divided into five parts:

(1) Socio-economic information and current travel behavior/ RP data,
(2) SP choice scenarios, including scenarios 1 and 2 only with travel mode choice, and
scenario 3 with travel mode choice as well as departure time choice when the trip is
flexible,
(3) Information services,
(4) Diagnostic questions,
(5) Attitudes and perceptions.
Current Travel Behavior
The first step is to collect the socio-economic information of the respondents, such as
individual characteristics (age, gender, occupation, education, and driver license),
household composition (children, teenagers and adults), income levels, residential
location, car ownership, parking availability and conditions, and transit pass ownership.
The respondents are then asked to recall all the trips that they have made during
yesterday or the last weekday, and provide origins, destinations, start time, end time,
transport modes, distance and purposes for all these trips (Revealed Preference data).
Figure 1 presents a sample webpage of the questions regarding the respondents’ socio-
economic information (in Portuguese).
March 2009 7
Figure 1. Webpage for Socio-Economic Information of the Respondents
Based on some sampling rules, we would choose one trip (with distance longer than 1
km) out of all the RP trips for each respondent. The respondents need to reply more
detailed information for the selected trips depending on the RP travel modes. For private
car, he/she should remember the parking fee, toll, transit pass ownership, schedule
flexibility, the number of people who shared the trip with him/her, and whether public
transports were available for the trip. For public transport, he/she should remember the
access/egress mode, the access/egress time, waiting time, the number of transfer, transit
fare, and transit pass ownership. For car and heavy mode, he/she should remember the
parking fee, toll, the access/egress time, waiting time, the number of transfers, transit pass
ownership, and whether other public transports were available for the trip.
SP Choice Scenarios
Based on the information of the selected RP trip (origin, destination, travel time, and
departure time), three hypothetical SP scenarios are presented to each respondent.
Scenarios 1 and 2 only provide a choice of transport modes. Scenario 3 provides a choice
of transport modes as well as a choice of departure time intervals when the scheduling of
the selected RP trip is flexible, and only a choice of transport modes when inflexible.
March 2009 8
As mentioned in section 1, a key challenge for this SP survey is the organization of the
large choice set of transport modes and their attributes. These alternatives are divided into
three groups: car-based modes (drive alone in a private car, one-way car rental, carpool,
shared taxi), public transportation (bus, heavy mode including train, subway and ferry,
minibus), and multi-transport modes (bus and heavy mode, park & ride (with child drop-
off), one-way car rental with heavy mode). Respondents are asked to select one
‘preferred’ mode from each group. The three preferred modes were presented to
respondents for a ‘final choice’ in a separate exercise. For example, a respondent may be
provided with the attribute values for car-related modes in Table 1, the attribute values
for public transportation in Table 2, and the attribute values for multi-transport modes in
Table 3. Suppose he selected private car, minibus, and bus and heavy modes as three
preferred modes. He would be provided with the same attributes values for these three
modes again in Table 4 and asked to make a final choice.
The availability of transport modes depends on the respondents’ car ownership, driver
license, RP trip origins and destinations, trip purposes, etc. Congestion charge only
applies to trips entering the congested areas of Lisbon during 07:00 to 20:00.
Table 1. An Example of Car-Based Alternatives and Their Attributes
Table 2. An Example of Public Transportation and Their Attributes
March 2009 9
Table 3. An Example of Multi-Transport Modes and Their Attributes
Table 4. An Example of Final Choice of Three Preferred Modes
In the third scenario, the respondents would be asked to choose their departure time if the
selected RP trip is flexible for scheduling, as shown in Table 5.
Table 5. An Example of Departure Time Choice
March 2009 10
Information Services
This section focuses on people’s preference among different information services,

including message boards, Internet-based, cell phone, and in-vehicle GPS navigation
systems. For example, Table 6 presents the attributes of new services.
Table 6. An Example of Preference for New Information Services
Diagnostic Questions
The following questions are used to verify that respondents understand the SP scenarios
and have made logical choices.
! Were you able to understand the choice scenarios as they were presented?
! In the choice scenarios, did you think the alternatives offered to you realistic?
! When considering the options which of the following factors did you consider?
Travel time/cost/convenience/flexibility
Attitudes and Perceptions
Respondents’ attitudes and perceptions of transport modes may affect their preference
and choices (see Outwater et al. 2003). Respondents were required to indicate their levels
of agreement with some statements (rank from 1 strongly disagree to 5 strongly agree),
such as
! I can count on the car to get me to my destination on time (reliability)

! The car offers me the flexibility I need for my schedule (flexibility)
! Using public transport is environmentally friendly (environment)
! Public transport is not secure (security)
! Public transport is very crowded (comfort)
March 2009 11
2.1.3 Experimental Design and Test
Hypothetical attribute values are generated for SP choice scenarios using experimental
design techniques. For transport mode choice, the minimal fractional factorial design
(SAS software) applies for each group of alternatives. Five levels were assumed for each
attribute. Attribute values also depend on the information of RP trip, including travel
time, cost, origin, destination, and transit pass ownership. After eliminating
extreme/dominant combinations, weights are assigned to valid combinations with lower
probabilities for the highest and lowest levels.
Table 7 presents the attribute values and levels for one-way car rentals as an example.
The values in parenthesis are the assigned weights/probabilities of each level.
Table 7. Attribute Values and Levels for One-Way Car Rental

Levels 0 (worst) 1 2 3 4 (best)
Time from door to door
1.5 (0.5) 1.2 (1) 1.0 (1) 0.8 (1) 0.5 (0.5)
(* current travel time)
Time variability
0.5 (0.5) 0.375 (1) 0.25 (1) 0.125 (1) 0.05 (1)
Fuel cost (* current
2.5 (0.5) 2 (1) 1 (1) 0.5 (1) 0.375 (0.5)
cost)
Congestion charge
5 (1) 3 (1) 2 (1) 1 (1) 0.5 (1)
(Euros)
Parking fee (Euros) 5 (0.5) 3 (1) 2 (1) 1 (1) 0.5 (1)
Rental cost (Euros) 15 (0.5) 10 (1) 8 (1) 5 (1) 3 (1)
Fare (* current travel
0.4 (0.5) 0.3 (1) 0.2 (1) 0.1 (1) 0.0625 (1)
time)
Waiting time
0.375 (0.5) 0.25 (1) 0.2 (1) 0.125 (1) 0.05 (1)
A Monte-Carlo simulation is used to verify the experimental design. In this process,

synthetic attribute values of alternatives are generated. A set of initial coefficients and
model structure are selected. Synthetic choices are generated next using these selected
coefficients and model structure. The models are then estimated with this full set of
synthetic data (attributes and generated choices).
The estimation results exhibited a close match between the initial coefficients and
estimated coefficients, which proved the validity of the experimental design.
2.2 Data
A pilot study of the SP survey was conducted early September 2008. It was used to test
the survey structure and to test the experimental design validity. There were 150
respondents, selected by the company that implemented the pilot survey; these
respondents were sampled in order to roughly correspond to the general socioeconomic
characteristics of the Lisbon Metropolitan Area population.
Each respondent was required to choose among transport modes for three SP scenarios.
There were a total of 450 observations of transport mode choice. In terms of departure
March 2009 12
time choice, there were a total of 71 observations for respondents who had flexibility in
trip scheduling.
2.2.1 General Findings
On average, respondents were able to complete the entire questionnaire around 20

minutes, as shown in Figure 2.
Figure 2. Distribution of Survey Time for Pilot Study
Except the 8 respondents (5% of total respondents) who have left the diagnostic questions
unanswered, all respondents reported that they could understand the SP scenarios. 94% of
all respondents felt that the alternatives offered were realistic, as shown in Figure 3.
Figure 3. Answers to Diagnostic Questions
March 2009 13
2.2.2 Respondent Characteristics
Here are some basic statistics about these respondents at the time of the survey,
! Most respondents were between the ages of 18 to 45
! 50% of the respondents were female
! 141 respondents were employed full-time, 7 were employed part-time, and 2 were
students
! A majority of the respondents were educated to the high-school or college level,
as shown in Figure 4
! Over 50% of the respondents’ household incomes were between 1000 and 2000
Euros per month, as shown in Figure 5
! Approximately 66% of the respondents had driver licenses, and 54% were car
owners
! Out of the 100 respondents who had driver licenses, most of them had parking
space inside building or used on-street parking space, as shown in Figure 6
Figure 4. Distribution of Education Levels
March 2009 14
Figure 5. Distribution of Household Income
Figure 6. Distribution of Parking Types Near Home
2.2.3 RP Mode Choice Data
In the pilot study, the purposes of most RP trips were either commuting to work (41%) or
returning to home (43%). Other trip purposes included commuting with intermediate
March 2009 15
stop, returning home with intermediate stop, business/other work related, shopping,
leisure/entertainment, and picking-up/dropping-off someone else, as shown in Figure 7.
The existing transport modes in Lisbon included solo driver, carpool, car as passenger,
bus, heavy mode, bus and heavy mode, car and heavy mode, and accommodate. Figure 8
presents the distribution of transport modes for the selected RP trips. Approximately 25%
of these RP trips were via car, and 75% were via public transportation. This can be
explained by the low amount of people with driver licenses, low car ownership, and
convenient coverage of bus routes and heavy modes in Lisbon.
Figure 7. Distribution of the Purposes for Selected RP Trips
March 2009 16
Figure 8. Distribution of Travel Modes for Selected RP Trips
2.2.4 SP Mode Choice Data
The SP choice scenarios were generated based on the selected RP trip of each respondent.
Figure 9 presents the share of travel modes, including current transport modes,
innovative transport modes and also new combinations of existing and innovative,
according to the results of ‘final choice’ of SP modes.
The total share of public transportation (bus, heavy modes, minibus, and bus and heavy
modes) was around 75% in the SP scenarios, which was almost equal to the share of
public transportation (bus, heavy modes, and bus and heavy modes) for selected RP trips.
That is to say, the innovative modes are more likely to compete with similar modes: e.g.
minibus vs. current public transport modes, rental car vs. private car, shared taxi vs.
private car.
Minibus (34%) has induced a lot demand from other modes, because its speed is designed
to be higher than bus and its cost is relatively lower than private car. This is consistent
with the good receptivity from the focus group. Shared taxi (5%) has induced certain
demand from private car, due to the lower fare and no need to search for parking space.
Since large percentage of the selected RP trips were to commute or return home, one-way
car rental (1%) was not so attractive in this case.
March 2009 17
Figure 9. Distribution of Travel Modes Based on the SP Final Choice
2.2.5 Departure Time Choice Data
In terms of departure time choice, there were a total of 71 observations of respondents

who had flexibility in trip scheduling. We assume that departure time after 19:00 pm not
available for the trips with purpose of commuting to work or commuting with
intermediate stop, and that departure time before 7:00 am and between 7:00-8:00 am are
not available for the trips with purpose of returning home or returning home with
intermediate stop.
Figure 10 and Figure 11 present the distribution of departure time for the selected RP
trips and for the SP choice respectively. The number of SP trips departing during 7:00-
8:00 pm was larger than that of RP trips, which conflicted with the purpose of peak
spreading of congestion pricing. This was carefully reviewed while updating the survey.
March 2009 18
Figure 10. Distribution of Departure Time for the Selected RP Trips
Figure 11. Distribution of Departure Time for the SP Choice
2.3 Model Estimation
Mode choice and departure choice models were estimated separately for the pilot data.
Different model specifications were compared in this regard. Each group of models is
March 2009 19
described below along with candidate variables, evaluated model structures and
parameter estimates.
In general, discrete choice models based on maximum likelihood estimation technique

were used for the models. These models capture the influence of attributes and
characteristics on the decision makers’ preferences (as shown in Figure 12).
Figure 12. Discrete Choice Framework
2.3.1 Mode Choice Model
Variables
The characteristics of the respondents and the trip in question considered for mode choice
model estimation are as follows:
! Age
! Employment status (full-time/part-time)
! Employment type (professional, associate)
! Education level
! Gender
! Household income
! Household size
! Number of children below 10 years in the household
! Number of children aged 11-17 years in the household
! Purpose of trip
! Flexibility of the trip
The attributes of the alternatives considered for mode choice model estimation are as
follows:
! Door-to-door travel time
! Travel time variability
March 2009 20
! Fuel cost (for private car and carpool)
! Congestion charge (for private car and carpool)
! Parking fee (for private car and carpool)
! Rental fee (for one-way car rental)
! Fare (for public transport modes)
! Access time (for public transport modes)
! Access time variability (for public transport modes)
! Waiting time (for public transport modes)
! Number of transfers (for public transport modes or multimodal options)
! Service fee (for park-and-ride with child drop-off facility)
It may be noted that not all attributes are applicable to all modes.
Another important attribute is the availability of the mode for the trip in question. Some
of the new smart modes are designed to cater certain origin-destination pairs and/or
certain times of day. For example: express minibuses are intended to serve sub-urban
commuters.
Estimated Models
Multinomial Logit (MNL) Model
Several model specifications were tested with the MNL structure. The specification
resulting the best goodness-of-fit value has been presented here. The utility functions for
this specification can be expressed as follows:
Car ASC_Car * one + BETA_TT_a * TT_Car + BETA_TV_a * TV_Car + BETA_TC_a *

TC_Car + BETA_CC * CC_Car + BETA_PF * PF_Car + BETA_AGE1830_a *
Age_1830 + BETA_AGE45_a * Age_45 + BETA_FULLTIME_a * FullTime +
BETA_INCOME2_a * Income_2 + BETA_HOUSEHOLDSIZE_a * HouseholdSize +
BETA_COMMUTE_a * Commute + BETA_RETURNHOME_a * ReturnHome +
BETA_DEPARTUREFLEXIBLE * Departure_Flexible + BETA_CHILDREN_a *
Children
Rental ASC_Rental * one + BETA_TT_a * TT_Rental + BETA_TV_a * TV_Rental +
BETA_TC_a * TC_Rental + BETA_CC * CC_Rental + BETA_AGE1830_a *
Age_1830 + BETA_AGE45_a * Age_45 + BETA_FULLTIME_a * FullTime +
BETA_INCOME2_a * Income_2 + BETA_HOUSEHOLDSIZE_a * HouseholdSize +
BETA_COMMUTE_a * Commute + BETA_RETURNHOME_a * ReturnHome +
BETA_DEPARTUREFLEXIBLE * Departure_Flexible + BETA_CHILDREN_a *
Children
Carpool ASC_Carpool * one + BETA_TT_a * TT_Carpool + BETA_TV_a * TV_Carpool +
BETA_TC_a * TC_Carpool + BETA_CC * CC_Carpool + BETA_PF * PF_Carpool +
BETA_AGE1830_a * Age_1830 + BETA_AGE45_a * Age_45 + BETA_FULLTIME_a
* FullTime + BETA_INCOME2_a * Income_2 + BETA_HOUSEHOLDSIZE_a *
HouseholdSize + BETA_COMMUTE_a * Commute + BETA_RETURNHOME_a *
ReturnHome + BETA_DEPARTUREFLEXIBLE * Departure_Flexible +
BETA_CHILDREN_a * Children
Taxi ASC_Taxi * one + BETA_TT_a * TT_Taxi + BETA_TV_a * TV_Taxi + BETA_TC_a *
TC_Taxi + BETA_AGE1830_a * Age_1830 + BETA_AGE45_a * Age_45 +
BETA_FULLTIME_a * FullTime + BETA_INCOME2_a * Income_2 +
BETA_HOUSEHOLDSIZE_a * HouseholdSize + BETA_COMMUTE_a * Commute +
March 2009 21
BETA_RETURNHOME_a * ReturnHome + BETA_DEPARTUREFLEXIBLE *
Departure_Flexible + BETA_CHILDREN_a * Children
BusWalk ASC_BusWalk * one + BETA_TT_b * TT_BusWalk + BETA_AT * AT_BusWalk +
BETA_TR_b * TR_BusWalk + BETA_TC_b * TC_BusWalk
HeavyWalk ASC_HeavyWalk * one + BETA_TT_b * TT_HeavyWalk + BETA_AT *
AT_HeavyWalk + BETA_TR_b * TR_HeavyWalk + BETA_TC_b * TC_HeavyWalk
Minibus ASC_Minibus * one + BETA_TT_b * TT_Minibus + BETA_AT * AT_Minibus +
BETA_AV * AV_Minibus + BETA_TR_b * TR_Minibus + BETA_TC_b * TC_Minibus
BusHeavy ASC_BusHeavy * one + BETA_TT_c * TT_BusHeavy + BETA_AT * AT_BusHeavy +
BETA_FR * FR_BusHeavy + BETA_TC_c * TC_BusHeavy
ParkRide ASC_ParkRide * one + BETA_TT_c * TT_ParkRide + BETA_AT * AT_ParkRide +
BETA_FR * FR_ParkRide + BETA_TC_c * TC_ParkRide + BETA_PF * PF_ParkRide
The explanations of the variables are presented in the following tables (Table 8 and Table
9) and the estimation results are presented in Table 10. Note that since none of the
respondents chose combination of one-way car rental and heavy mode in the data, the
alternative was removed from the choice set during model estimation.
Table 8. Explanations of of Socio-Economic Variables

Socio-economic
Attibute meanings
variables
Age_1830 Dummy variable, 1 only if respondent’s age >= 18 and <= 30
Age_45 Dummy variable, 1 only if respondent’s age > 45
FullTime Dummy variable, 1 only if respondent is full-time employed
Dummy variable, 1 only if respondent’s monthly household income is greater than or
Income_12
equal to1000 euros and also less than 2000 euros
Dummy variable, 1 only if respondent’s monthly household income is greater than or
Income_2
equal to 2000 euros
HouseholdSize Number of people in respondent’s household
Dummy variable, 1 only if the trip purpose is commute to work or commute with
Commute
intermediate stop
Dummy variable, 1 only if the trip purpose is return home or return home with
ReturnHome
intermediate stop
Departure_Flexible Dummy variable, 1 only if the respondent indicates he/she has flexible departure time
Dummy variable, 1 only if the respondent has children age <= 10 years old in
Children_10
household
Dummy variable, 1 only if the respondent has childrean age between 11 and 17 years
Children_1117
old in household
Dummy variable, 1 only if the respondent has children <= 10 years or between 11-17
Children
years old in household
Female Dummy variable, 1 only if the respondent is female
Dummy variable, 1 only if the respondent’s occupation is professional, associate
Professional
professional and technical occupation, or mangers and senior officials
Worker Dummy variable, 1 only if the respondent’s occupation is skilled worker
HighSchool Dummy variable, 1 only if the respondent’s education level is high school
Dummy variable, 1 only if the respondent’s education level is college or post-
College
graduate
March 2009 22
Table 9. Explanation of Variables Related to Attributes Transport Modes
Car-based Modes
Drive alone in a
Attributes One-way car rental Carpool Taxi
private car
Availabilty of travel
Car_Available Rental_Available Carpool_Available Taxi_Available
mode
Time from door to
TT_Car TT_Rental TT_Carpool TT_Taxi
door
Time variability TV_Car TV_Rental TV_Carpool TV_Taxi
Fuel cost FC_Car FC_Rental FC_Carpool
Congestion charge CC_Car CC_Rental CC_Carpool
Parking fee PF_Car PF_Carpool
Rental cost RC_Rental
Taxi fare TF_Taxi
Waiting time WT_Taxi
Travel cost (except TC_Rental =
TC_Carpool = TC_Taxi =
congestion charge TC_Car = FC_Car FC_Rental +
FC_Carpool TF_Taxi
and parking fee) RC_Rental
Public Transport Modes
Bus (access Heavy mode (access
Attribute Minibus
walking) walking)
BusWalk_Available HeavyWalk_Available Minibus_Available
mode
Time from door to
TT_BusWalk TT_HeavyWalk TT_Minibus
door
Time variability TV_BusWalk TV_HeavyWalk TV_Minibus
Access time AT_BusWalk AT_HeavyWalk AT_Minibus
Access time
AV_Minibus
variability
Waiting time WT_BusWalk WT_HeavyWalk WT_Minibus
Transfers TR_BusWalk TR_HeavyWalk TR_Minibus
Transit fare TF_BusWalk TF_HeavyWalk TF_Minibus
Travel cost (except
TC_BusWalk = TC_HeavyWalk = TC_Minibus =
congestion charge
TF_BusWalk TF_HeavyWalk TF_Minibus
and parking fee)
March 2009 23
Table 9. Explanation of Variables Related to Attributes Transport Modes (cont.)
Multi-modal Alternatives
Park & Ride for heavy
Heavy mode (access
Attributes Bus and heavy mode mode or Park & Ride with
one-way car rental)
children drop-off
BusHeavy_Available ParkRide_Available HeavyRental_Available
mode
Time from door to door TT_BusHeavy TT_ParkRide TT_HeavyRental
Time variability TV_BusHeavy TV_ParkRide TV_HeavyRental
Access time AT_BusHeavy AT_ParkRide AT_HeavyRental
Frequency FR_BusHeavy FR_ParkRide FR_HeavyRental
Transfers TR_BusHeavy TR_ParkRide TR_HeavyRental
Transit fare TF_BusHeavy TF_ParkRide TF_HeavyRental
SP_ParkRide (only when
respondent needs to drop-off
Service price chidren <10 and the trip SP_HeavyRental
purpose is commuting with
an intermediate stop)
Parking fee PF_ParkRide PF_HeavyRental
TC_ParkRide =
TF_ParkRide or
TC_ParkRide =
Travel cost (except TF_ParkRide + TC_HeavyRental =
TC_BusHeavy =
congestion charge and SP_ParkRide (only when TF_HeavyRental +
TF_BusHeavy
parking fee) respondent needs to drop-off SP_HeavyRental
children < 10 and the trip
purpose is commuting with
an intermediate stop)
March 2009 24
Model: Multinomial Logit
Number of estimated parameters: 30
Number of observations: 450
Number of individuals: 435
Init log-likelihood: -849.845
Final log-likelihood: -589.564
Adjusted rho-square: 0.270968
Final gradient norm: 0.00250252
Table 10. Estimation Results of MNL Model for Transport Mode Choice
Coefficient Value Robust Std err Robust t-test
ASC_Car 0.727 0.883 0.824*
ASC_Rental -2.62 1.14 -2.31
ASC_Carpool -0.581 0.853 -0.681*
ASC_BusWalk 0 (fixed)
ASC_HeavyWalk -0.104 0.188 -0.551*
ASC_Minibus 0.759 0.227 3.34
ASC_BusHeavy -0.973 0.458 -2.13
ASC_ParkRide -1.58 0.581 -2.71
BETA_TT_a -0.0401 0.00856 -4.68
BETA_TT_b -0.0430 0.00616 -6.98
BETA_TT_c -0.0272 0.00640 -4.25
BETA_TV_a -0.0419 0.0135 -3.11
BETA_AT -0.0311 0.0107 -2.90
BETA_AV -0.0281 0.0239 -1.17*
BETA_TC_a -0.103 0.0556 -1.85*
BETA_TC_b -0.918 0.118 -7.77
BETA_TC_c -0.596 0.122 -4.90
BETA_CC -0.707 0.522 -1.36*
BETA_PF -0.254 0.147 -1.72*
BETA_TR_b -0.00238 0.146 -0.0163*
BETA_FR -0.0225 0.0380 -0.593*
BETA_DEPARTUREFLEXIBLE 0.618 0.329 1.88*
BETA_AGE1830_a -0.0220 0.414 -0.0531*
BETA_AGE45_a 0.677 0.442 1.53*
BETA_CHILDREN_a 0.319 0.279 1.14*
BETA_COMMUTE_a -1.42 0.617 -2.30
BETA_RETURNHOME_a -1.47 0.601 -2.44
BETA_FULLTIME_a -1.38 0.526 -2.63
BETA_HOUSEHOLDSIZE_a 0.0963 0.0319 3.02
BETA_INCOME2_a 1.039 0.338 3.08
Generally, the signs of all coefficients are consistent with our priori assumptions.
Negative signs for travel time and cost means that an increase in travel time or cost will
reduce the utility (and thus the chosen probability) of an alternative. If all other attributes
are equal, car and minibus are the most preferable choices for the respondents.
March 2009 25
The travel time variability emerges as a significant attribute for car-related modes. This is
because cars are highly sensible to traffic congestion. Accurate predications of travel time
with information services would increase car use. For car-related modes, the sensitivities
to different forms of cost were found to be as follows:
congestion charge > parking fee > travel cost
This indicates that introducing the congestion charge and increasing the parking price is
likely to have a greater impact in deterring car usage compared to increasing the general
travel cost (e.g. fuel cost, rental cost etc.). Dummy variable of departure time flexibility is
has positive coefficient, therefore people who prefer flexibility is more likely to choose
car-related modes.
Waiting time is not a well-defined attribute for public transportation compared with
frequency, and its coefficient is insignificant. The number of transfers is also
insignificant. These were noted in updating the main survey.
Significant socio-economic variables include the age of the respondent, number of

children in the respondents' household, trip purpose, employment status, household size,
and monthly household income.
Nested Logit (NL) Model
Nested Logit (NL) models were also applied for transport mode choice, assuming there
were three nests for transport modes: car-based nest (drive alone in private car, one-way
car rental, carpool, shared taxi), public transportation nest (bus, heavy mode, minibus),
and multi-modal nest (bus and heavy mode, park & ride with child drop-off, one-way car
rental with heavy mode). The best model specification was same as that of the selected
MNL model. The estimation results of the NL model are shown in Table 11.
Firstly, the coefficients of the three nests were set to one. The top coefficient for the NL
structure was estimated with u = 0.58, which was different from zero or one. The
conclusions are similar to those of the selected MNL model. The main differences are
that the constant of ASC_car is negative and the sign of the coefficient for respondents
aged between 18 and 30 is positive in the NL model.
Model: Nested Logit

March 2009 26
Table 11. Estimation Results of NL Model for Transport Mode Choice
ASC_Car -0.484 2.45 -0.198*
ASC_Rental -4.24 2.88 -1.47*
ASC_Carpool -1.87 2.36 -0.791*
ASC_Taxi -1.59 2.55 -0.623*
ASC_BusWalk 0 (fixed)
ASC_HeavyWalk -0.130 0.200 -0.650*
ASC_Minibus 0.682 0.267 2.56
ASC_BusHeavy -2.52 2.07 -1.22*
ASC_ParkRide -2.98 1.98 -1.50*
BETA_TT_a -0.0412 0.0133 -3.10
BETA_TT_b -0.0559 0.0141 -3.96
BETA_TT_c -0.0299 0.00868 -3.44
BETA_TV_a -0.0411 0.0177 -2.32
BETA_AT -0.0409 0.0174 -2.35
BETA_AV -0.0297 0.0289 -1.03*
BETA_TC_a -0.0964 0.0717 -1.35*
BETA_TC_b -1.24 0.326 -3.80
BETA_TC_c -0.781 0.215 -3.64
BETA_CC -0.855 0.787 -1.09*
BETA_PF -0.340 0.198 -1.71*
BETA_TR_b -0.0410 0.160 -0.257*
BETA_FR -0.0192 0.0541 -0.356*
BETA_DEPARTUREFLEXIBLE 0.820 0.569 1.44*
BETA_AGE1830_a 0.0775 0.676 0.115*
BETA_AGE45_a 1.11 0.899 1.23*
BETA_CHILDREN_a 0.484 0.473 1.02*
BETA_COMMUTE_a -2.87 2.19 -1.31*
BETA_RETURNHOME_a -2.69 1.87 -1.44*
BETA_FULLTIME_a -1.63 0.843 -1.94*
BETA_HOUSEHOLDSIZE_a 0.145 0.0783 +1.86*
BETA_INCOME2_a 1.97 1.35 +1.46*
U 0.582 0.303 1.92 (0 test)
-1.38* (1 test)
2.3.2 Departure Time Choice Model
For departure time choice, two types of attributes were generated: early or late schedule
delay and log interval size.
! Early and late schedule delay

For time interval (a, b), if the respondent’s revealed departure time dt >= b, he/she when
choosing time interval (a, b) would tend to depart at a time close to b. This results in an
early schedule delay ESD = dt – b. If dt < a, he/she when choosing time interval (a, b)
would tend to depart at a time close to a, resulting in a late schedule delay LSD = a –dt. If
otherwise a <= dt < b, then there are no early schedule delay or late schedule delay when
he/she chooses this time interval. For example, a respondent actually departed at 8:30 am
for a commute trip to work. If he chooses time interval (7:00, 8:00), he would face an
early schedule delay equal ESD_mp = 8:30 – 8:00 = 0.5 hour. If he chooses time interval
(8:00, 10:30), he can still depart at his usual time 8:30 am and there is no early schedule
delay or late schedule delay. If he chooses time interval (10:30, 19:00), he would face a
March 2009 27
late schedule delay equal LSD_a = 10:30 – 8:30 = 2 hours. We define that the early
schedule delay of time interval after 19:00 pm equals zero (ESD_e = 0), and the late
schedule delay of time interval before 7:00 am equals zero (LSD_em = 0).
! Log interval size

The variable of the Log interval size is used to describe the size of each chosen time
interval. Since the chosen probability of the time interval is proportional to the size of the
time interval, the utility function of the Multinomial Logit (MNL) model should contain a
variable for the Log interval size and the corresponding coefficient need to be constrained
to one. We assume that the time interval of a trip departing before 7:00 am is 4 hours
from 3:00 am to 7:00 am, and the time interval of a trip departing after 19:00 pm is 8
hours from 19:00 pm to 3:00 am next day.
The explanations of variables used for departure time choice are listed in Table 12.
Regarding to departure time choice, we had total 71 observations since only respondents
with flexible trip scheduling were required to make such choice. It was much less than
the number of observations (450) for transport mode choice, and the data quality was
under concern. In order to avoid unreasonable and confusing results, we did not include
socio-economic variables in the MNL model specifications. The estimation results of the
best model for departure time choice are presented in Table 13.
EM ASC_em * one + BETA_TT_ac * TT_em + BETA_ESD_I * ESD_em_I + BETA_ESD_II *

ESD_em_II + BETA_FC_ac * FC_em + BETA_PF_ac * PF_em + BETA_IN * IN_em
MP ASC_mp * one + BETA_TT_b * TT_mp + BETA_ESD_I * ESD_mp_I + BETA_ESD_II *
ESD_mp_II + BETA_LSD_I * LSD_mp_I + BETA_LSD_II * LSD_mp_II + BETA_FC_b *
FC_mp + BETA_CC_b * CC_mp + BETA_PF_b * PF_mp + BETA_IN * IN_mp
M ASC_m * one + BETA_TT_ac * TT_m + BETA_ESD_I * ESD_m_I + BETA_ESD_II *
ESD_m_II + BETA_LSD_I * LSD_m_I + BETA_LSD_II * LSD_m_II + BETA_FC_ac * FC_m
+ BETA_CC_c * CC_m + BETA_PF_ac * PF_m + BETA_IN * IN_m
A ASC_a * one + BETA_TT_ac * TT_a + BETA_ESD_I * ESD_a_I + BETA_ESD_II * ESD_a_II
+ BETA_LSD_I * LSD_a_I + BETA_LSD_II * LSD_a_II + BETA_FC_ac * FC_a +
BETA_CC_c * CC_a + BETA_PF_ac * PF_a + BETA_IN * IN_a
E ASC_e * one + BETA_TT_ac * TT_e + BETA_LSD_I * LSD_e_I + BETA_LSD_II * LSD_e_II
+ BETA_FC_ac * FC_e + BETA_PF_ac * PF_e + BETA_IN * IN_e
March 2009 28
Table 12. Attribute Notations of Departure Time
Departure Between 7:00- Between 8:00- Between
Attribute meanings After 19:00
before 7:00 8:00 10:30 10:30-19:00
Availability of
EM_Available MP_Available M_Available A_Available E_Available
departure time
Time from door to
TT_em TT_mp TT_m TT_a TT_e
door (in mins)
Time variability (in
TV_em TV_mp TV_m TV_a TV_e
mins)
Fuel cost FC_em FC_mp FC_m FC_a FC_e
Congestion charge CC_mp CC_m CC_a
Parking fee PF_em PF_mp PF_m PF_a PF_e
Early schedule delay
ESD_em ESD_mp ESD_m ESD_a ESD_e
(in hours)
Late schedule delay
LSD_em LSD_mp LSD_m LSD_a LSD_e
(in hours)
Log interval size IN_em IN_mp IN_m IN_a IN_e
ESD_em_I ESD_mp_I ESD_m_I ESD_a_I ESD_e_I
(the part <= 5 hours)
ESD_em_II ESD_mp_II ESD_m_II ESD_a_II ESD_e_II
(the part > 5 hours)
Late schedule delay
LSD_em_I LSD_mp_I LSD_m_I LSD_a_I LSD_e_I
(the part <= 2 hours)
Late schedule delay
LSD_em_II LSD_mp_II LSD_m_II LSD_a_II LSD_e_II
(the part > 2 hours)
Model: Multinomial Logit

Null log-likelihood: -87.5813
Likelihood ratio test: 72.3173
March 2009 29
Table 13. Estimation Results of MNL Model for Departure Time Choice
ASC_em 0 (fixed)
ASC_mp 5.10 1.36 3.74
ASC_m 1.77 0.798 2.22
ASC_a 1.69 1.33 1.26*
ASC_e 1.08 1.65 0.658*
BETA_TT_ac -0.0297 0.0126 -2.35
BETA_TT_b -0.0212 0.0474 -0.447*
BETA_ESD_I -0.458 0.255 -1.80*
BETA_ESD_II 0.638 0.226 2.82
BETA_LSD_I -0.993 0.579 -1.71*
BETA_LSD_II -0.683 0.396 -1.72*
BETA_FC_ac -1.05 0.278 -3.79
BETA_FC_b -1.48 0.797 -1.86*
BETA_CC_b -1.01 0.671 -1.50*
BETA_CC_c -0.318 0.432 -0.735*
BETA_PF_ac -0.627 0.166 -3.78
BETA_PF_b -2.38 0.606 -3.94
BETA_IN 1 (fixed)
The values of most coefficients are consistent with our priori assumptions. Trips initiated
during the morning peak are more sensitive to travel time due to traffic congestion.
People are more sensitive to late schedule delays than early schedule delays, because
tardiness usually results in negative consequences.
For morning peak hours, the cost sensitivity has following relationship: parking fee > fuel
cost > congestion charge. While for the other four intervals, the cost sensitivity: fuel cost
> parking fee > congestion charge. These conflict with the conclusions drawn from the
modeling of transport mode choice (the cost sensitivity to congestion charge > parking
fee > travel cost). The observations of departure time choice were of smaller sample and
poor quality. Therefore, the conclusions from transport mode choice are more reliable.
2.4 Attitudes to Car and Public Transportation
Respondents’ attitudes to the current transport system may affect their preference for
different transport modes. Latent variables, which are indicated with the responses to
attitudinal questions, can be used in the modeling.
In the pilot study, most respondents agreed that cars provided comfort, good privacy and
security, but were not environmentally friendly and needed latest technology to provide
sufficient traffic information, as shown in Figure 13.
According to Figure 14, most respondents agreed that the current system of public
transportation in Lisbon provided good service, comfort, reliability, schedule flexibility,
and better time use, and was clean and environmentally friendly. It had short travel, low
travel cost, short access and egress time due to good coverage.
March 2009 30
Figure 13. Car-Related Statements with Obvious Agreement
March 2009 31
Figure 14. Public Transport-Related Statements with Obvious Agreement
March 2009 32
Figure 14. Public Transport-Related Statements with Obvious Agreement (cont.)
March 2009 33
Figure 15. Statements without Obvious Agreement or Disagreement
There were no obvious agreement or disagreement with following statements: car is

reliable, schedule flexible, has short travel time and low travel cost; car users should pay
more for traffic congestion; public transport is crowded, as shown in Figure 15.
March 2009 34
2.5 Information Services
The section of information services is for the use of CityMotion: Data Fusion for
Mobility Consumers, Providers, and Planners (another project of MIT-Portugal
Initiative). Part of the results is summarized here.
Figure 16 presents the usage of current information services. Message boards (63%) were
widely used in Lisbon to provide traffic guidance. There were some market shares for
Internet-based information service (25%) and GPS navigation system (12%). However,
cell phone was found to be rarely used.
Figure 16. Usage of Current Information Services
March 2009 35
3 Revised Stated Preference Survey
3.1 Main Adjustments from Pilot Study
An important objective of pilot study was to test the SP survey structure and the validity
of experimental design. According to the estimation results, there were much fewer and
lower quality responses to the SP scenarios of departure time choice compared with those
to the SP scenarios of travel mode choice; some of the attributes of innovative modes
were not found to be as important as expected, e.g., sensitivity to waiting time and
number of transfers were insignificant. Furthermore, average survey time was longer than
20 min so that the SP questionnaire needed to be shortened.
Substantial changes were made to solve these problems and especially to get a more
robust and rich SP survey. The key changes in the survey structure are as follows:
! Integrated the departure time choice with the mode choice exercises to better
capture the effect of congestion charge
! Reduced the number of SP scenarios to two
! Eliminated the section on information services from the survey
In addition, the context of SP survey was adjusted as follows:

! Deleted the alternative of carpool but added regular taxi
! Included preferred occupancy as a choice dimension to better represent car
sharing
! Used time-varying attributes, such as travel time, congestion charge, frequency
and access time to better represent the context of departure time choice
! Added new questions and attributes (parking fee and search time for parking
space) in SP scenarios regarding to parking pricing and enforcement
! Added regular fee as an attribute, if freeway was tolled (charged)
! Waiting time replaced by frequency
In Section 3.2, we present the updated SP scenarios. Several changes were made in the
experimental design of the survey as well, which are documented in Section 3.3.
3.2 Updated SP Scenarios with Multidimensional Choice
The first dimensional choice of SP scenarios was the mode choice, the second one the
departure time choice and the third one occupancy choice (if applicable).
In the revised survey, the new travel modes here included shared taxis, express minibus,
one-way car rentals, park-and-ride systems with a tutored delivery of children to their
schools, and one-way car rentals with heavy mode. Alongside the five existing modes
(car, regular taxi, bus, heavy mode, and bus and heavy modes), this yielded a travel mode
choice set of up to ten alternatives per respondent. The first dimensional choice of SP
scenarios was the choice consisted of these existing and new travel modes.
The level-of-service of these alternatives varied substantially on the departure time; in

particular there were significantly long travel time and high costs (in the form of
March 2009 36
congestion charge and parking enforcement) for traveling in peak hours. This is expected
to strongly influence the individual travel pattern and the choice of departure time
intervals was included in the SP survey as a second dimension.
In addition, it is expected that these radically different modes and level-of-service are
likely to foster the sharing of trips. A third dimension has been added in the choice
structure: the choice of occupancy for private car, one-way car rentals and regular taxi.
The organization and presentation of a multidimensional choice exercise with such a

large number of alternatives and attributes posed a challenge for the revised SP survey,
since cognitive burden may force respondents to adopt a simple decision strategy based
on only partial information (Caussade et al. 2005). To minimize the complexity, a similar
grouping as in the Pilot study was implemented: the alternatives were first presented
sequentially in three groups: car-based, public transport based and multi-modal (see
Table 14, Table 15 and Table 16). Respondents were asked to choose their ‘preferred’
combinations of mode, departure time and occupancy (if applicable) in each case. The
three preferred combinations were then presented in a single choice task and the
respondent was asked to select the combination of his ‘choice’. In this organization, each
respondent was presented with two SP scenarios, four choice tasks, at most four
alternatives per choice task, and at most seven attributes per alternative (see Table 17).
For an instance of the multidimensional choice for car-based group (see Table 14), here
are the descriptions for some attributes.
! The time from door to door includes the time spent to reach the car/taxi, the actual
driving time, the time spent to reach the final destination after getting out of the
car/taxi, and the waiting time (especially for shared taxi). This time may vary
from day to day depending on traffic conditions. The +/- sign in the time
represents this variability. The time may be more predictable (less variation) if
real-time travel information is available.
! The fuel cost, congestion charge, parking fee, regular toll and other costs are the
total amount of out-of-pocket costs for your trip, regardless of whether or not you
share the expenses with household members, co-workers, neighbors or others. We
assume that if respondents receive reimbursement for toll/parking, that these
reimbursements are still valid.
! The travel time, congestion charge, and waiting time of shared taxi may vary
depending on the departure time and is the highest during peak periods. Peak
periods are 8:00 to 10:30 in the morning and 16:30 to 20:00 in the afternoon.
! Mean time to find a parking spot is the time used to search for an available
parking spot near your final destination. If parking enforcement is strict, it means
when you park illegally you will certainly be fined or your car will be towed.
! Preferred occupancy refers to the possible total number of people among whom
the costs will be shared (either formally or informally). Respondents are not
required to choose preferred occupancy for shared taxi, since the respondents
cannot predict how many people will share the taxi.
March 2009 37
Table 14. An Example of Multidimensional Choice for Car-Based Group
Features Private car One way car rental Regular Taxi Shared Taxi
Before 7:00: 15 +/- 2 Before 7:00: 15 +/- 2 Before 7:00: 15 +/- 2 Before 7:00: 15 +/- 2
min min min min
7:00-8:00: 20 +/- 5 min 7:00-8:00: 25 +/- 5 min 7:00-8:00: 20 +/- 5 min 7:00-8:00: 25 +/- 5 min
8:00-10:30: 35 +/- 10 8:00-10:30: 40 +/- 10 8:00-10:30: 40 +/- 10 8:00-10:30: 45 +/- 15
Time from
min min min min
door to door
10:30-12:00: 20 +/- 5 10:30-12:00: 25 +/- 5 10:30-12:00: 20 +/- 5 10:30-12:00: 25 +/- 5
- driving time,
min min min min
access time,
12:00-16:30: 20 +/- 5 12:00-16:30: 25 +/- 5 12:00-16:30: 20 +/- 5 12:00-16:30: 25 +/- 5
egress time, and
min min min min
waiting time
16:30-20:00: 35 +/- 10 16:30-20:00: 40 +/- 10 16:30-20:00: 40 +/- 10 16:30-20:00: 45 +/- 15
min min min min
After 20:00: 15 +/- 2 After 20:00: 15 +/- 2 After 20:00: 15 +/- 2 After 20:00: 15 +/- 2
min min min min
Fuel cost 3 Euros - - -
Before 7:00: no charge Before 7:00: no charge
Congestion
7:00 to 8:00: 1 Euros 7:00 to 8:00: 1 Euros
charge
8:00 to 10:30: 2 Euros 8:00 to 10:30: 2 Euros
- the fee you
10:30-12:00: 1 Euros 10:30-12:00: 1 Euros - -
should pay for
12:00-16:30: 1 Euros 12:00-16:30: 1 Euros
entering central
16:30-20:00: 1 Euros 16:30-20:00: 1 Euros
Lisbon
After 20:00: no charge After 20:00: no charge
Rental cost (including
Additional Parking fee: 1 Euros fuel cost and parking
Fare: 8 Euros Fare: 5 Euros
costs Regular toll: 0.5 Euro fee): 5 Euros
Regular toll: 0.5 Euro
Waiting time: Waiting time:
Before 7:00: 2 +/- 1 Before 7:00: 2 +/- 1
min min
7:00-8:00: 5 +/- 2 min 7:00-8:00: 3 +/- 1 min
Mean time to find a 8:00-10:30: 8 +/- 3 min 8:00-10:30: 5 +/- 3 min
parking spot: 5 min 10:30-12:00: 5 +/- 2 10:30-12:00: 3 +/- 1
Other
Parking enforcement: min min
strict 12:00-16:30: 5 +/- 2 12:00-16:30: 3 +/- 1
min min
16:30-20:00: 8 +/- 3 16:30-20:00: 5 +/- 3
min min
After 20:00: 2 +/- 1 min After 20:00: 2 +/- 1 min
Before 7:00 Before 7:00 Before 7:00 Before 7:00
7:00 to 8:00 7:00 to 8:00 7:00 to 8:00 7:00 to 8:00
Preferred 8:00 to 10:30 8:00 to 10:30 8:00 to 10:30 8:00 to 10:30
Travel Mode 10:30 to 12:00 10:30 to 12:00 10:30 to 12:00 10:30 to 12:00
and Departure 12:00 to 16:30 12:00 to 16:30 12:00 to 16:30 12:00 to 16:30
Time 16:30 to 20:00 16:30 to 20:00 16:30 to 20:00 16:30 to 20:00
After 20:00 After 20:00 After 20:00 After 20:00
1 people 1 people 1 people
Preferred 2 people 2 people 2 people
Occupancy 3 people 3 people 3 people
>=4 people >=4 people >=4 people
Regarding to parking pricing and enforcement, attributes such as parking fee and search
time were introduced in choice scenarios to evaluate travelers’ responses and sensitivity
March 2009 38
(see Hensher and King 2001, Alberta and Mahalel 2006). In addition, travelers were
asked about their perception of parking conditions at/near trip destinations and their
personal attitudes to parking problems as follows.
Which of the following are applicable to parking facilities at/near your destination (select
all that are applicable):
Parking is fully free on the street
There are some paid and some free parking spots on the street
There are dedicated parking lots
Parking is strongly and effectively enforced
It is easy to find a parking spot
Please indicate your level of agreement with the following statements (rank from 1
strongly disagree to 5 strongly agree).
Difficulty in getting a parking spot near the destination is the main problem of using
the car.
Parking illegally is a major offence.
High parking cost is a major problem of using the car.
Table 15. An Example of Multidimensional Choice for Public Transportation

Bus Heavy mode
Features Minibus
(access by walking) (access by walking)
Time from door Before 7:00: 35 +/- 2 min Before 7:00: not available
to door 7:00-8:00: 35 +/- 2 min 7:00-8:00: not available
- includes in- 8:00-10:30: 50 +/- 5 min 8:00-10:30: 40 +/- 2 min
vehicle travel 10:30-12:00: 35 +/- 2 min 45 min +/- 1 min 10:30-12:00: not available
time access time, 12:00-16:30: 35 +/- 2 min 12:00-16:30: not available
egress time, and 16:30-20:00: 50 +/- 5 min 16:30-20:00: 40 +/- 2 min
waiting time After 20:00: 35 +/- 2 min After 20:00: not available
Access time By walking: 5 min By walking: 10 min By walking: 10 min
Before 7:00: 15 min Before 7:00: 10 min Before 7:00: not available
7:00-8:00: 15 min 7:00-8:00: 10 min 7:00-8:00: not available
8:00-10:30: 10 min 8:00-10:30: 5 min 8:00-10:30: 30 min
Frequency 10:30-12:00: 15 min 10:30-12:00: 10 min 10:30-12:00: not available
12:00-16:30: 15 min 12:00-16:30: 10 min 12:00-16:30: not available
16:30-20:00: 10 min 16:30-20:00: 5 min 16:30-20:00: 30 min
After 20:00: 15 min After 20:00: 10 min After 20:00: not available
Transfers 2 1 -
Transit 3 Euros (without pass) or 2 Euros (without pass) or
2 Euros
Fare/Pass 0.5 Euros (with pass) 0.5 Euro (with pass)
Before 7:00 Before 7:00
7:00 to 8:00 7:00 to 8:00
Preferred 8:00 to 10:30 8:00 to 10:30
8:00 to 10:30
Travel Mode 10:30 to 12:00 10:30 to 12:00
16:30 to 20:00
and Departure 12:00 to 16:30 12:00 to 16:30
Time 16:30 to 20:00 16:30 to 20:00
After 20:00 After 20:00
March 2009 39
Table 16. An Example of Multidimensional Choice for Multi-Modal Group
Park & Ride for Heavy One way car rental with
Bus and Heavy mode
Features mode Heavy mode (access/egress
(access/egress by bus)
(access by car) by one way car rental)
Time from door
to door
- includes in-
vehicle travel 50 min +/- 5 min 40 min +/- 5 min 45 min +/- 5 min
time access time,
egress time, and
waiting time
By bus: By driving: By driving:
Before 7:00: 5 +/- 2 min Before 7:00: 8 +/- 2 min Before 7:00: 8 +/- 2 min
7:00-8:00: 5 +/- 2 min 7:00-8:00: 10 +/- 2 min 7:00-8:00: 10 +/- 2 min
8:00-10:30: 10 +/- 5 min 8:00-10:30: 15 +/- 5 min 8:00-10:30: 15 +/- 5 min
Access time
10:30-12:00: 5 +/- 2 min 10:30-12:00: 10 +/- 2 min 10:30-12:00: 10 +/- 2 min
12:00-16:30: 5 +/- 2 min 12:00-16:30: 10 +/- 2 min 12:00-16:30: 10 +/- 2 min
16:30-20:00: 10 +/- 5 min 16:30-20:00: 15 +/- 5 min 16:30-20:00: 15 +/- 5 min
After 20:00: 5 +/- 2 min After 20:00: 8 +/-2 min After 20:00: 8 +/- 2 min
Before 7:00: 10 min Before 7:00: 10 min Before 7:00: 10 min
7:00-8:00: 10 min 7:00-8:00: 10 min 7:00-8:00: 10 min
Frequency
8:00-10:30: 5 min 8:00-10:30: 5 min 8:00-10:30: 5 min
- Level of
10:30-12:00: 10 min 10:30-12:00: 10 min 10:30-12:00: 10 min
Service for
12:00-16:30: 10 min 12:00-16:30: 10 min 12:00-16:30: 10 min
Heavy mode
16:30-20:00: 5 min 16:30-20:00: 5 min 16:30-20:00: 5 min
After 20:00: 10 min After 20:00: 10 min After 20:00: 10 min
Transfers 2 1 1
Transit 3 Euros (without pass) or 1 Euros (without pass) or 1 Euros (without pass) or
Fare/Pass 0.5 Euros (with pass) 0.5 Euros (with pass) 0.5 Euros (with pass)
Fuel cost: 2 Euros
Parking fee: 1 Euros
Rental cost (including fuel
Additional costs - Service price: 2 Euros
cost and parking fee): 3 Euros
Supervised by: school
teachers
Before 7:00 Before 7:00 Before 7:00
7:00 to 8:00 7:00 to 8:00 7:00 to 8:00
Preferred 8:00 to 10:30 8:00 to 10:30 8:00 to 10:30
Travel Mode 10:30 to 12:00 10:30 to 12:00 10:30 to 12:00
and Departure 12:00 to 16:30 12:00 to 16:30 12:00 to 16:30
Time 16:30 to 20:00 16:30 to 20:00 16:30 to 20:00
After 20:00 After 20:00 After 20:00
Service for
Yes
Children Drop-
No
off by Tutors
Assuming a respondent told us that in the previous three situations, he preferred to use
private car during 7:00-8:00 with the occupancy of 2 people, to take minibus during 8:00-
10:30, and to take bus and heavy mode during 8:00-10:30. The three preferences were
unchangeable and fixed with the same values as before. He would be asked to make a
final choice among his three preferences, as shown in Table 17. Time-varying attributes
were only presented in the table with the values of preferred departure times chosen in
the previous three situations.
March 2009 40
Table 17. An Example of Final Choice from Three Preferences
Private Car/7:00-8:00/ Bus and Heavy mode/8:00-
Features Minibus/8:00-10:30
Occupancy of 2 people 10:30 (access/egress by bus)
Time from door
to door
- includes in-
vehicle travel 7:00-8:00: 20 +/- 5 min 8:00-10:30: 40 +/- 2 min 50 min +/- 5 min
time access time,
egress time, and
waiting time
Frequency: Frequency:
Level of Service - 8:00-10:30: 30 min 8:00-10:30: 5 min
Transfers: 0 Transfers: 2
Fuel cost: 3 Euros 3 Euros (without pass) or
Cost Transit fare: 2 Euros
Parking fee: 1 Euros 0.5 Euros (with pass)
Congestion charge:
Additional Cost 7:00 to 8:00: 1 Euros - -
Regular toll: 0.5 Euro
Mean time to find a
Access time by walking: Access time by bus:
Others parking spot: 5 min
10 min 8:00-10:30: 10 +/- 5 min
Parking enforcement: strict
Final Choice of
Travel Mode
together with
Departure Time
and Occupancy
Due to the variety of travelers’ characteristics and trip information, the appearance of SP
scenarios could vary slightly. For example, if a respondent does not have a car in the
household, the alternative of private car would not be presented to him/her as a possible
choice; if a respondent makes a trip with origin and destination both outside Lisbon,
congestion charge that applies to trips entering Lisbon would not be presented as an
attribute to affect his/her choice.
In the revised SP survey, the availabilities of travel modes posed complexity issues as in
the Pilot survey. In addition, and attributes were explicitly defined. The availability rules
are presented in Appendix A.
3.3 Updates in the Experimental Design
Based on the pilot study, the share of minibus was much larger than expected, while the
innovative modes of one-way car rental and park & ride with children drop-off were not
so attractive as expected. These might be affected by the purposes of the selected trips
and the deficiencies in experimental design of attribute levels, e.g., the travel times of
minibus were relatively shorter than proposed value and the advantages of one-way car
rental were not emphasized.
In the experimental design for the revised SP survey, more explicit rules were applied to
generate the attribute levels. To uniform the whole design, standard SP travel times were
March 2009 41
calculated based on the travel times, departure times and travel modes of the selected RP
trips. The design of attribute values, such as time from door to door, time variability,
access time and transfer, was separated when the standard SP travel times <=15 min, 15-
30 min, 30-45 min, and > 45 min. Furthermore, there were some rules that captured the
inter relationships among travel times and access times of different modes, among costs
(e.g. fuel cost and rental cost) and travel times.
Fractional factorial design was used and elimination rules were then proposed to refine
these initial outcomes, e.g., to delete combinations with a dominate alternative, and to
delete combinations with too large differences among travel times and costs of different
modes.
March 2009 42
4 Future Plans and Challenges
Generally speaking, the practical design of a SP survey under complex and multiple
scenarios is an extremely time-consuming process. Trials and errors are needed to
generate artificial but close to reality choice scenarios and attribute values. This large-
scale SP survey conducted in Lisbon, Portugal is remarkable and provides a nice example
for future applications of SP survey.
Up to date, the design for the revised SP survey and experimental exercises has been
completed. This survey will be conducted using Internet and computer-assisted personal
interviews (CAPI). The programming of SP survey is now at the final stage. The Internet
survey will be implemented and data will be collected during March and May in 2009
using mailing lists and divulgation through newspaper and websites in Lisbon, and the
CAPI will be conducted during May and June in 2009 to correct the sampling biases.
The choice scenarios of the revised SP survey were more complicated and robust than
before, which poses challenges to the modeling and estimation in future. There are two
problems under discussion: how to model multidimensional choice, and how to deal with
a large choice set.
Although Multinomial Logit (MNL) models are commonly applied for discrete choice
analysis, it is not suitable for the case of multidimensional choice. By virtue of the fact,
the alternatives in a multidimensional choice set share observed and also unobserved
attributes along various dimensions. There exists a significant amount of literature
focusing on the modeling techniques, such as joint logit model and nested logit model for
destination and mode choice (Ben-Akiva and Lerman 1985), multinomial probit model
for brand choice (Raap and Franses 2000), mixed multinomial logit and ordered logit
model for residential location and car ownership decision (Bhat and Guo 2007), error
components logit model for time-of-day and mode choice (De Jong et al. 2003), and
mixed logit model for vehicle choice (Hess et al. 2006). Different models need to be
applied and compared for the revised SP survey data with three-dimensional choice of
travel mode, departure time and occupancy.
Furthermore, all the alternatives of the large multidimensional choice set were classified
based on the 1st dimension of travel mode into three groups with similarities: car-based
group, public transportation group and multi-modal group. In order to reduce cognitive
burden, each respondent was only asked to select the preferred travel mode together with
departure time and occupancy in each group per time. Then, three preferred alternatives
were assumed unchangeable and presented to the respondent again for a final choice.
Though there has been research on approaches to deal with large choice sets in consumer
choice settings, e.g., telecom features (Ben-Akiva and Gershenfeld 1998), magazine
subscription (McAlister 1979), entertainment services (Venkatesh and Mahajan 1993),
and auto-ownership (Hanson and Martin 1990), to our knowledge, there has not been
significant research about how to deal with travel choice context where presenting a large
number of alternatives is essential given the particular scenarios of application.
March 2009 43
The specific organization of alternatives raised a number of methodological issues. We
can explore the following questions in future research.
! Are ‘preference’ and ‘choice’ data inherently different?
! Can they be combined in a consistent manner?
! In the combined data, does the a priori assumption about the nesting structure (the
grouping of alternatives used in the survey) still govern?
A possible approach is to compare the estimation results of a particular model with only
the ‘choice’ responses and only the ‘preference’ responses against the estimate results of
the pooled model (where both ‘preference’ and ‘choice’ responses are considered). A
scale parameter needs to be introduced in the pooled model to account for the probable
difference in variance. The model performance can be compared later against a model
where separate individual specific error terms for ‘preference’ and ‘choice’ responses are
tested to capture intra-respondent and inter-respondent heterogeneities (Bliemer et al.
2008, Louviere et al. 2008, Rose et al. 2008).
March 2009 44
References
Alberta, G. and Mahalel, D. (2006). Congestion Tolls and Parking Fees: A Comparison
of the Potential Effect on Travel Behavior. Transport Policy, Vol. 13, No. 6, pp. 496-502.
Ben-Akiva, M. and Gershenfeld (1998). Multi-featured Products and Services: Analyzing

Pricing and Bundling Strategies. Journal of Forecasting, Vol. 17, pp. 175-196.
Ben-Akiva, M. and Lerman, S.R. (1985). Discrete Choice Analysis: Theory and
Application to Travel Demand. Cambridge, MA: MIT Press.
Bhat, C.R. and Castelar, S. (2002). A Unified Mixed Logit Framework for Modeling
Revealed and Stated Preferences: Formulation and Application to Congestion Pricing
Analysis in the San Francisco Bay Area. Transportation Research Part B, Vol. 36, No. 7,
pp. 593-616.
Bhat, C.R. and Guo, J.Y. (2007). A Comprehensive Analysis of Built Environment
Characteristics on Household Residential Choice and Auto Ownership Levels.
Transportation Research Part B, Vol. 41, pp. 506-526.
Bliemer, M., Rose, J. and Hensher, D.A. (2008). Constructing Efficient Stated Choice
Experiments Allowing for Differences in Error Variances Across Subsets of
Alternatives. Transportation Research B, forthcoming.
Caussade, S., Ortuzar, J.D., Rizzi, L.I. and Hensher, D.A. (2005). Assessing the Influence
of Design Dimensions on Stated Choice Experiment Estimates. Transportation Research
Part B, Vol. 39, pp. 621-640.
De Jong, G., Daly, A., Pieters, M., Vellay, C., Bradley, M. and Hofman, F. (2003). A
Model for Time of Day and Mode Choice Using Error Components Logit.
Transportation Research Part E, Vol. 39, pp. 245-268.
Hanson, W.A. and Martin, R.K. (1990). Optimal Bundle Pricing. Management Science,
Vol. 36, pp. 155-174.
Hensher, D.A. (2006). How Do Respondents Process Stated Choice Experiments?

Attribute Consideration Under Varying Information Load. Journal of Applied
Econometrics, Vol. 21, pp. 861-878.
Hensher, D.A. and King, J. (2001). Parking Demand and Responsiveness to Supply,
Pricing and Location in the Sydney Central Business District. Transportation Research
Part A, Vol. 35, No. 3, pp. 177-196.
Hess, S., Train, K.E. and Polak, J.W. (2006). On the Use of a Modified Latin Hypercube
Sampling (MLHS) Method in the Estimation of a Mixed Logit Model for Vehicle Choice.
Transportation Research Part B, Vol. 40, pp. 147-163.
March 2009 45
Louviere, J.J., Street, D., Burgess, L., Wasi, N., Islam, T. and Marley, A.A.J. (2008).
Modeling the Choices of Individual Decision-makers by Combining Efficient Choice
Experiment Designs with Extra Preference Information. Journal of Choice Modeling,
Vol. 1-1, pp. 128-163.
Outwater, M.L., Castleberry, S., Shiftan, Y., Ben-Akiva, M., Zhou, Y.S. and Kuppam, A.
(2003). Attitudinal Market Segmentation Approach to Mode Choice and Ridership
Forecasting: Structural Equation Modeling. Transportation Research Record, Vol. 1854,
pp. 32-42.
Raap, R. and Franses, P.H. (2000). A Dynamic Multinomial Probit Model for Brand
Choice with Different Long-run and Short-run Effects of Marketing-mix Variables.
Journal of Applied Econometrics, Vol. 16, pp. 717-744.
Rose, J., Bliemer, M.C., Hensher, D.A. and Collins, A.T. (2008). Designing Efficient
Stated Choice Experiments in the Presence of Reference Alternatives. Transportation
Research Part B, Vol. 42, pp. 395-406.
Sanko, N. (2001). Guidelines for Stated Preference Experiment Design, Master Thesis,
Ecole Nationale des Ponts et Chaussees, School of International Management, Research
project in association with RAND Europe.
Venkatesh, R. and Mahajan, V. (1993). A Probabilistic Approach to Pricing a Bundle of

Products or Services. Journal of Marketing Research, November, pp. 494-508.
Viegas, J.M., de Abreu e Silva, J. and Arriaga, R. (2008). Innovation in Transport Modes
and Services in Urban Areas and their Potential to Fight Congestion. Presented at the 1st
Annual Planning Conference on Planning Research, Evaluation in Planning, Porto,
Portugal, May 30th.
March 2009 46
Appendix A
Availability of Transport Modes

Private car
! Only available for respondents who have a car in the household (driver license not
required for passengers), or
! Chose Private car or Car and Heavy mode for their selected RP trip, or
! Mentioned that they have access to a car for the trip even if they chooses public
transportation for selected RP trip
One-way car rental
! Only available for respondents who have driver license or whose household
members have driver license (as passenger), or
! Chose Private car or Car and Heavy mode for the selected RP trip
! Only available for certain origin-destination pairs with this innovative service
! Not available for trips with purposes such as commute to work, commute to
school, or commute with intermediate stop
Regular taxi
! Available for every respondent
Shared taxi
! Available for every respondent
! Not available for trips with purposes such as commute to work, commute to
school, or commute with intermediate stop
Bus
! Available for every respondent, given that the current bus service in Lisbon has
good coverage
Heavy mode
! Only available for certain origin-destination pairs with good coverage of heavy
mode
Minibus
! Only available for certain origin-destination pairs with this innovative service
! Only available for trip with purpose such as commute to work, commute to
school, commute with intermediate stop, return home or return home with
intermediate stop
! Only available during peak hours 8:00-10:30 and 16:30-20:00
Bus and Heavy mode
! Available for every respondent, given the current bus service in Lisbon has good
coverage and the bus routes and heavy mode stations are well connected
Park & Ride for heavy mode
! Use the same rules as private car
One-way car rental with Heavy mode
! Use the same rules as one-way car rental
March 2009 47
Availabilities of Attributes
Time-dependent attributes, such as time from door to door, congestion charge, frequency
and access/egress time
! For a trip with purpose such as commute to work, commute to school, or
commute with intermediate stop, only the alternatives and attribute values for the
first four time intervals will be shown to the respondent (before 7:00, 7:00-8:00,
8:00-10:30, and 10:30-12:00). But if the revealed departure time of this trip is
after 12:00, all alternatives and attribute values for seven time intervals will be
presented to the respondent.
! For a trip with purpose such as return home or return home with intermediate
stop, only the alternatives and attribute values for the last four time intervals will
be shown to the respondent (10:30-12:00, 12:00-16:30, 16:30-20:00, and after
20:00). But if the revealed departure time of this trip is before 10:30, all
alternatives and attribute values for seven time intervals will be presented to the
respondent.
Congestion charge
! Only appears for selected RP trip with origin outside Lisbon and destination
inside Lisbon, i.e., trip entering congested area of Lisbon.
! For residents inside Lisbon, there are discounted charges of 80% for their trips
entering Lisbon.
Parking fee
! If the last trip of the day is chosen and the trip purpose is return home or return
home with intermediate stop, there is no parking fee required (residential parking
space).
! If the last trip of the day is chosen and the trip purpose is not return home or
return home with intermediate stop (with very small probability), we should
discard the last trip of this respondent and choose another RP trip of this
respondent, which is better to calculate the parking fee and duration.
Regular toll
! This is the toll for using freeway. This value is fixed and generated based on the
RP trip origin-destination.
Transit pass
! The single-trip cost of transit pass is calculated based on its monthly cost, which
is around the half cost of transit fare. Its availability is based on the answers
regarding to the ownership of transit pass.
! For trip with purpose such as commute to work, commute to school, or commute
with intermediate stop, even if the respondent does not have transit pass now, we
assume that he will buy and use transit pass in future and the price of transit pass
is presented hear instead of the price of single-trip fare.
Service price and supervised by school teachers or professional people
! Only appear here when park & ride for heavy mode is available for the
respondent, the trip purpose is commute with intermediate stop, and the
respondent has at least one child in household less than 10 years old who needs to
be drop-off.
March 2009 48

Stated Preference Surveyfor New Smart Transport Modesandservicesdesignpilotstudyandnewrevision

Uploaded by

Copyright:

Available Formats

Stated Preference Surveyfor New Smart Transport Modesandservicesdesignpilotstudyandnewrevision

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Stated Preference Surveyfor New Smart Transport Modesandservicesdesignpilotstudyandnewrevision

Uploaded by

Copyright:

Available Formats

Transportation Systems

Working Paper Series

Stated Preference Survey for New Smart

Lang Yang, Charisma F Choudhury & Moshe Ben-Akiva

João Abreu e Silva & Diana Carvalho

MIT Portugal Program

Lang Yang ([email protected]), Charisma F Choudhury ([email protected]),

Intelligent Transportation Systems Lab

João Abreu e Silva ([email protected]), Diana Carvalho ([email protected])

Instituto Superior Tecnico

The objective of SCUSSE (Smart Combination of Passenger Transport Modes and

! One-way car rental

! Park and ride with child drop-off

New traffic management methods include the following:

Motivation of Stated Preference Survey

! To evaluate acceptability of innovative modes and services

Challenges and Complexity

Furthermore, new smart services introduced in Lisbon include congestion pricing,

The process of developing the SP survey is summarized in seven steps:

Structure of This Paper

The remainder of this paper is organized as follows.

Section 4 talks about future plans and modeling challenges.

2.1 Development of the SP Survey

2.1.1 Focus Group Discussion

2.1.2 Framing the Questions of the SP Survey

The large questionnaire of SP survey was divided into five parts:

Current Travel Behavior

Table 1. An Example of Car-Based Alternatives and Their Attributes

Table 2. An Example of Public Transportation and Their Attributes

Table 4. An Example of Final Choice of Three Preferred Modes

Table 5. An Example of Departure Time Choice

This section focuses on people’s preference among different information services,

Table 6. An Example of Preference for New Information Services

Attitudes and Perceptions

! I can count on the car to get me to my destination on time (reliability)

Table 7. Attribute Values and Levels for One-Way Car Rental

A Monte-Carlo simulation is used to verify the experimental design. In this process,

2.2.1 General Findings

On average, respondents were able to complete the entire questionnaire around 20

Figure 2. Distribution of Survey Time for Pilot Study

Figure 3. Answers to Diagnostic Questions

Figure 4. Distribution of Education Levels

Figure 6. Distribution of Parking Types Near Home

2.2.3 RP Mode Choice Data

Figure 7. Distribution of the Purposes for Selected RP Trips

2.2.4 SP Mode Choice Data

2.2.5 Departure Time Choice Data

In terms of departure time choice, there were a total of 71 observations of respondents

Figure 11. Distribution of Departure Time for the SP Choice

2.3 Model Estimation

In general, discrete choice models based on maximum likelihood estimation technique

Figure 12. Discrete Choice Framework

2.3.1 Mode Choice Model

Multinomial Logit (MNL) Model

Car ASC_Car * one + BETA_TT_a * TT_Car + BETA_TV_a * TV_Car + BETA_TC_a *

Table 8. Explanations of of Socio-Economic Variables

congestion charge > parking fee > travel cost

Significant socio-economic variables include the age of the respondent, number of

Nested Logit (NL) Model