Modelling of Transport Mode Selection by Travellers to Perth City

1
Modelling of Transport Mode Selection by
Travellers to Perth City

• Sharif Siddique
• PATREC Research Forum
• 3 October 2005

2
(No Transcript)
3
Presentation Summary

• Background
• Objective of this study
• Characteristics of the sample
• Method used
• Modeling
• Results
• Conclusion

4
Background

• PhD topic
• Joint development of Genetic Algorithm and a
  Stated Preference survey to optimise policies for
  ameliorating the environmental impacts of
  transport

5
Background

• Study area is Perth city
• Air quality analysis
• CO and NOx models
• Identifying policy measures to control air
  pollution
• Estimating the impacts of measures on air quality

6
Background

• Assessing travel mode choice behaviour of
  travellers to Perth city.
• Assessing the reactions of car users on policy
  measures using stated preference survey.
• Formulating an optimal composite environment
  and transport policy using evolutionary approach.

7
This study

• Mainly focuses to assess travellers travel mode
  selection behaviour.
• Considers only those whose destination is the
  Perth city.
• Estimates travel demand elasticities with respect
  to trip cost.
• Estimates of value of travel time savings (VTTS).

8
Descriptive of the sample
9
Some more descriptive
10
Choice alternatives

• Initial investigation
• Number of car
• Licence status
• Age
• Train service availability
• Developing discrete choice nested logit model

11
Method used

• Discrete choice method
• The purpose for reviewing travellers choice of
  transport mode is to examine the relationship
  between the mode-choice decision and a set of
  explanatory variables.
• The discrete choice model is the most popular
  method for analysis mode choice.

12
Discrete Choice Model

• Piq Pr (Uiq gt Ujq) probability of q
  individual choosing alternative i
• Where, Ui Vi ei V is observed
  component and e is unobserved component

13
Use of discrete choice model

• Elasticity estimation
• VTTS estimation

?
14
Data imputation

• Travel time for car is assumed to be same from
  the same suburb with similar distance.
• Travel times for bus and train are estimated
  using the Transperths web site
• Trip cost for car is assumed to be only the fuel
  cost. Petrol is considered only fuel used.
• A standard MultiRider 40 rate is used to
  estimate bus and train fares.

15
The Model

• Variable choice set is used
• CSET 1, means only Bus
• CSET 2, means EITHER (Car as passenger and Bus)
  OR (Bus and Train)
• CSET 3, means EITHER (Car as driver, Car as
  passenger, Bus) OR (CarB, Bus, Train)
• CSET 4, means Car as driver, Car as passenger,
  Bus, Train
• Explanatory variables used
• Modal attributes
• Travel time and trip cost
• Socio-demographic variables
• No. of vehicles, Parking fee, Personal income,
  Age, Gender

16
Utility functions used
UcarD ?carD ?tt Traveltime ?tc Tripcost
?carD vehicle ?carD parking UcarP
?carP ?tt Traveltime ?tc Tripcost
?carP vehicle ?carPage UBus ?Bus
?tt Traveltime ?tc Tripcost ?Busincome
?Bus sex UTrain ?tt Traveltime
?tc Tripcost
17
Models investigation

• 4 models are developed
• Multinomial logit (MNL) model without demographic
  variables
• MNL model with demographic variables
• Nested logit (NL) model
• NL model with different travel time parameters

models
18
back
MNL structure
NL structure
19
Models
20
Elasticity estimation

• Direct and cross-elasticities with respect to
  trip cost

21
Value of Travel Time Savings

• VTTS
• 5.70 per hour
• People like to save one hour of travel time in
  exchange of 5.70

22
Conclusion

• Mode selection behaviour
• Mode switching behaviour with respect to cost
• Trading off value for reducing travel time

Thank You