Content
• Quick Review of Major Concepts from Last Week
Accessibility (continued)
&
Basics of Travel Demand
– Accessibility measures via Gravity Model and Utility-based
Model
• Conclusions from Accessibility Lecture
– “Composite Measures,” Deciding on a “Best Measure”,
Accessibility as raison d’etre?
• Travel Demand
–
–
–
–
–
Day 6
11.953
Basic Characteristics
Primary Drivers Influencing Factors
International Comparisons
Implications for the Future…
• Assignment I
• Other Course Logistic Items
Back to the “Four Step”
Gravity-based Measures
• Theoretical origins in physics,
• Improvement over distance-based measures, partly
because they attempt to better reflect travel behavior
realities through their functional form, generally:
Ai = ∑ W j f (ci j , β )
j
Trip Generation
Predicts number of trips produced and attracted in a given zone
Trip Distribution
Produces trip production and attraction for each zone
Modal Split
• where:
–
–
–
–
Data Inputs
Inventories and forecasts of population, land uses, travel behavior, etc.
Predicts mode share typically for auto and public transport (can include walk, bike)
Wj represents the opportunities available in a given zone j;
f(cij, β) = exp (- βcij) = impedance between zones i and j;
cij represents the travel cost/distance between zones i and j; and
β is a travel cost sensitivity parameter.
• generally enters as a negative exponential function
• the accessibility measure clearly is highly sensitive to this parameter.
• Should come from empirical analysis
Utility-Based Accessibility: the Logit Model
Trip Assignment
Assigns trips to their respective networks
System Outputs
Provides, for each link, data including traffic volumes, speeds, vehicle mix
Evaluation
Back to the “Four Step”
Data Inputs
Inventories and forecasts of population, land uses, travel behavior, etc.
Pn (i) =
e μVin
j
∑e
j =1
μV jn
Example: Car or Bus?
• Potential Influencing
factors (variables)
– In-vehicle travel time
– Out-of-vehicle travel time
– Traveler income
– Age – Gender
– Etc.
Normally, Results used to MAKE PREDICTIONS about choices in
some future (or alternative)setting
Trip Generation
Predicts number of trips produced and attracted in a given zone
Trip Distribution
Produces trip production and attraction for each zone
Modal Split
Predicts mode share typically for auto and public transport (can include walk, bike)
Trip Assignment
Assigns trips to their respective networks
System Outputs
Provides, for each link, data including traffic volumes, speeds, vehicle mix
Evaluation
1
Utility-Based Accessibility: The “Logsum” and Nested Logit
Pn(dm) = Pn(m|d)Pn(d)
“Logsum” at “the root” represents composite
benefit (“Expected Maximum Utility”) of the
entire choice process
e(Vd +
V ' d ) μ
Pn (d ) =
L 2. Destination Choice
Disturbance term = εd
Scale parameter = μd
d
∑ e(Vd +V 'd ) μ
Traffic
Analysis
Zone
(TAZ)
d
d ∈D n
V 'd =
1
μm
ln
∑ e
(Vm +Vdm ) μ m
d1
d2
d3
L 1. Mode Choice
Disturbance term = εdm
Scale parameter = μm
m∈M nd
e(Vm +
Vdm ) μ
m
Pn (m | d ) =
m1
∑ e(Vm +Vdm ) μ
m
m2
m3
m∈
M nd
Social Accessibility Levels
Female Adult, Evaluated at Mean Relevant
Characteristics for Income Category
High Income
Middle Income
Low Income
“Utility-based” Measures
• Theoretically appealing
– Basis in behavioral theory and welfare economics
• Not immediately and easily convertible into
meaningful and understandable units
– Convertible into currency, time, but cumbersome
• Assumes utility linear with respect to income
– Nonpresence of income effect
• Still travel-biased measures
– Cannot immediately account for non trip-based
accessibility (e.g., not traveling; trip-chaining)
Relative Decline in Recreational
Accessibility
Middle Income Female
Loss of Auto
Loss of Bike
Loss of Metro
“Composite” or “Activity-based”
Approaches
• Essentially merging person-based (timespace) with utility-based
• Aims to account for people’s activities
throughout the day.
• Directly linked to “activity-based” travel
research
– Reflect activity re-scheduling, work-at-home
possibilities, etc.
• Data and computationally intensive
2
Activity-Based Example: Long-Term
Impacts of Congestion Pricing
“Best” Measure?
• No universally-agreed upon criteria
• An “ideal” accessibility measure should reflect:
– Different preferences among people,
– Scarcity of people’s time and money,
– Range of relevant travel (“impedance”) characteristics
• safety, convenience, comfort, aesthetics, etc.;
– Range of destination (“opportunity”) characteristics:
• safety, convenience, aesthetics, diversity, etc.
– Relevant traveler characteristics
• vehicle availability, age, disability status, etc.
– And be “operational,” interpretable, easily communicated.
Activity (ABA) versus Trip (TBA)
- Total effects lower (mean and median) for ABA
- ABA accounts for shifting to non-peak, change in activity pattern
(e.g., work at home)
¾ The composite, activity-based approach approaches
the theoretical ideal.
See, e.g.: Ramming, 1994; Bhat et al, 2000;
Handy and Clifton, 2001; Geurs and van Wee, 2004
Dong et al, 2005.
Accessibility: Indicator or Variable?
Accessibility as LUT raison d’etre?
• Examples here have shown accessibility as Indicator
– US Cities accessibility
– Neighborhood variation (Limanond and Niemeier, 2003)
– Total User Benefits (Martinez and Araya, 2000)
• Accessibility also used as variable (input)
– As determinant of some behavior or activity, influencing, e.g., residential
choice, mode choice, vehicle ownership, etc.
– Household’s worker(s) commute time(s) influencing residential choice
– distance to bus stops;
– “neighborhood accessibility”;
– “transit accessibility”;
– number of jobs within certain driving distance;
– distance to CBD;
– employment density within certain radii;
– number of establishments within various radii of home
– even population density or share of commercial space reflect inherent relative
nearness of people, stores, etc.
• The mobility-for-accessibility perspective implies a
largely utilitarian perspective
– we travel to derive accessibility (e.g., “travel is a derived demand”)
• But, travel is not always a “means” to an “end,”
– “travel liking” (due to adventure, variety, independence desires,
etc.) and not just for leisure trips, but for routine trips and not just
for auto use (see Ory and Mokhtarian, 2005)
– Extra travel as a means of “information gain” (i.e., better
information on products, space, etc.) (Arentze and Timmermans,
2005)
• Travel’s role in social class formation
– E.g., Vasconcellos (1997) details the role of the car in the
“making of the middle class.”
• And, of course, in combination: e.g., in integrated LUT models
Basics of Mobility Demand
Relevant Basic Characteristics
• Purposes:
– Work, Shopping, Social, Recreational, Business, School, Others
• Origin:
– e.g., Home-based work, Home-based school, etc. nonhome-based
shopping, etc.
• Stage:
– e.g., Stage 1, 2, etc.
• Mode:
– car, bus, rail, etc.
• Time of Day:
– e.g., AM-Peak, Off-peak, etc.
• Tour:
– combination of trips taken between “anchors” (activity-based
modeling); multiple activities in a single tour = “trip chaining”
Sources of Data
• Fundamental source
– Household Origin-Destination Survey
– Should be calculated for a given Metropolitan Area
• National-level Surveys
– E.g., Censuses
– In US: NPTS: 1969, 1977, 1983, 1990, 1995, 2001
(NHTS)
– 2001 NHTS: 26,000 households, national-level, 24­
hour “travel day” diary, plus 28-day “Travel Period” for
long-distance travel (see nhts.ornl.gov)
• Distance, Time, other?
3
Santiago (2001)
How do people get around?
US MSAs 2000: Work Trip Mode Share
Walk
Carpool
Saturday
Summer Weekday
40
Mode Share All Trips
Public
Transport
Weekday
Sunday
45
35
30
25
20
15
10
5
k
er
s
O
th
W
al
C
Source: Hanson, 2004.
Bi
ke
ct
iv
o
xi
ol
e
ro
Ta
s
M
et
Bu
Au
to
0
Drive
Solo
Range of “Developing World” Cities
Image removed for copyright purposes.
See Figure 9 in "mobility 2001." World
Business Council for Sustainable Development, 2001, p.18.
Why do we Travel: Trip Purposes?
Other
Comm ute
Social
http://www.wbcsd.org/web/projects/mobility/english_overview.pdf
Shop
Pers onal
Source: U.S. DOE, 2004.
Source: Various, see WBCSD, Table A-1.
Travel Demand: Relevant Personal
Choices
• Activity choices
– result in the number of tours and trips made
by a person for a certain purpose
• Destination choices
• Mode choices
– car, train, bus, tram, metro, etc.
• Time-of-day choices
• Route choices
Travel Demand (Activity Choice + Destination Choice + Mode Choice) = f (Socio-Economics/Demographics, Communication Patterns/Time Routines, Travel Costs (generalized), Modal Availability, Land Use Patterns)
4
Major Socio-Economic and
Demographic Drivers
The Stylized “S-Curve"
• Household Income
– Car ownership
Auto Ownership
• although elasticities different at different income
levels: e.g., S-curve
– Longer and more trips
– Higher demand for speed
• higher value of time
Threshold income
for auto ownership
Personal Income
National Motorization Rate
Where’s the S-Curve?
Household-Based S-Curves…
Autos and Motorized Two-Wheelers in Chennai, India (1993)
Figure removed for copyright purposes.
See Willoughby, Christopher. "Managing Motorization."
World Bank Report TWU-42, April 2000, p. 8.
http://www.worldbank.org/html/fpd/transport/publicat/twu_42.pdf
Percent of Households
70
Cars
60
Two Wheelers
50
40
30
20
10
0
<20
20
41
62
83
104 125 146 167 188 209 >209
Monthly Income (US$)
Source: Willoughby, 2000, p. 8.
Source: RITES, 1995
U.S. Household Travel:
Sated Demand?
0.900
12,000
0.800
0.700
0.600
8,000
0.500
6,000
0.400
0.300
4,000
2,000
Vehicle-miles per
capita
0.200
Vehicles per capita
0.100
19
50
19
60
19
70
19
80
19
86
19
88
19
90
19
92
19
94
19
96
19
98
20
00
20
02
0
0.000
Trips/Capita/Day
10,000
Vehicles per Capita
Vehicle Miles Per Capita
Income and the Demand for Trips
Personal
Business
Work/
School
Social
Social
Shopping
Shopping
Shopping
Work/
School
Work/
School
Work/
School
Income
Source: U.S. DOE, 2004.
5
Income and the Demand for Trips
Income and Mode Share - Santiago
80
3
% of All Trips
2.5
2
1.5
Walking
70
Auto
60
Public Transport
50
40
30
20
1
10
0.5
0
0
<$230
$230$455
$455$910
$910$1640
$1640$3300
>$3300
<
17
$1
17
$1
8
20
-$
08
$2
Income and Mode Share - São Paulo
94
$4
0
75
-$
50
$7
60
11
-$
0
16
$1
-$
65
28
>
5
86
$2
Income, Motorization Rate & Mode Share
– Santiago
Walking
60%
Auto
Public Transport
Auto Mode Share
60%
% of All Trips
4
49
-$
Source: SECTRA, 1991.
Source: Companhia do Metropolitano de São Paulo, 1999.
70%
16
$3
Monthly Incom e (US$ 1991)
Average Household Income (US$1997)
80%
16
-3
50%
40%
30%
20%
50%
400
Auto Mode Share
Veh/1000
350
300
40%
250
30%
200
150
20%
100
10%
50
0%
0
10%
Vehicles per 1000 Pop.
Trips/Person/Day
São Paulo 1997
0%
<230
230-455
455-910
910-1640
1640-3300
>3300
Monthly Income (US$1997)
Source: Companhia do Metropolitano de São Paulo, 1999.
Source: SECTRA, 1991.
The Theory of Constant Travel
Budgets
What Does Schafer Say About
This?
Hours/Day
Spent Traveling
United States (1995)
Netherlands (1995)
Norway (1992)
Australia (1986)
United States (1977)
Netherlands (1985)
Norway (1985)
Switzerland (1994)
Switzerland (1984)
Great Britain (1995)
W. Germany (1976)
W. Germany (1989)
Japan (1992)
Great Britain (1976)
Singapore (1991)
Singapore (1991)
Chana Villages (1988)
Tanzania VIII. (1986)
0
1
Travel Cost, Percent of
Disposable Income
20
Source: Updated data based on Schafer (1998).
10
Kilometers
Traveled/Person/Day
20 0
40
80
Figure by MIT OCW.
6
How Would Schafer Respond?
Passenger Travel Growth, Past 50 Years.
12,000
0.900
WED
100,000
EEU
PAO
CPA
10,000
SAS
LAM
MEA
AFB
1,000
0.800
10,000
0.700
8,000
0.500
6,000
0.400
4,000
2,000
PAS
World
0.600
0.300
Vehicle-miles per
capita
0.200
Vehicles per capita
0.100
0
0.000
19
50
19
60
19
70
19
80
19
86
19
88
19
90
19
92
19
94
19
96
19
98
20
00
20
02
Passenger km/Capita
FSU
Vehicle Miles Per Capita
NAM
Vehicles per Capita
Implies: Increased
Income=Increased Travel
100
100
1,000
10,000
100,000
Source: U.S. DOE, 2004.
GDP/Capita, SUS(1985)
Source: Updated data based on Schafer (1998).
Figure by MIT OCW.
But, be careful with National to
Global Level Averages…
7