State of the Art & Practice in ABM: How Can We Better Address Planning Needs and Unique Regional Conditions in NY?
Peter Vovsha, PhD
Principal,
PB Americas, New York
NYMTC, June 8, 2011 1
Shift in Paradigm
35 large MPOs (1 million +) in US: Half of them have developed or are developing ABM
Others plan ABM in future
All large-scale model development projects for the last 5 years were ABMs
State-wide strategic decisions to move to ABM: California
Florida
Ohio
Michigan
2NYMTC, June 8, 2011
CT-RAMP Family of ABMs
Coordinated Travel Regional Activity-based Modeling Platform
Main features: Explicit intra-household interactions and
Coordinated DAP (CDAP)
Continuous temporal dimension
Integration of activity generation, location, and TOD sub-models
JAVA-based package for ABM construction
3NYMTC, June 8, 2011
Members of CT-RAMP Family 1st generation:
Columbus, OH (MORPC) – in practice since 2004 Lake Tahoe, NV (TMPO) – in practice since 2006
2nd generation: Atlanta, GA (ARC) – in practice since 2009 San-Francisco Bay Area, CA (MTC) – in practice since 2010
3rd generation: San-Diego, CA (SANDAG) – started in 2008 Phoenix, AZ (MAG) – started in 2009 Jerusalem, Israel (JTMT) – started in 2009 Chicago, IL (CMAP) – started in 2010
Every model has many unique features
4NYMTC, June 8, 2011
CT-RAMP Basic Structure
NYMTC, June 8, 2011 5
Other Regional ABMs in Practice
Individual Daily Pattern: San-Francisco County, CA (SFCTA) – in practice since 2001 Sacramento, CA (SACOG) – in practice since 2006 Tel-Aviv, Israel (NETA) – in practice since 2009 Denver, CO (DRCOG) – in practice since 2010
Interactive Tour Generation: New York, NY (NYMTC) – in practice since 2002
Individual Daily Pattern with some Joint Travel Components Seattle, WA (PSRC) – started in 2008 Houston, TX (HGCOG) – started in 2011
Continuous Duration: Los-Angeles, CA (SCAG) – started in 2009
Population Synthesizer: Baltimore, MD (BMC) – started in 2011
State-wide (Intercity) Models: Oregon – in practice since 2008 Ohio – in practice since 2009
NYMTC, June 8, 2011 6
ABMs in the United States
NY
San Francisco
Seattle
ColumbusDenver
Atlanta
Sacramento
Bay Area
Developed by PB
Developed by others
Oregon
Ohio
San Diego
Lake Tahoe
CT-RAMP Family
PhoenixLA
Chicago
7NYMTC, June 8, 2011
Houston
Baltimore
What Can we Learn from These ABMs?
Selected model features relevant to BPM 2.0: Limitations of BPM 1.x
Unique conditions of NY
Forthcoming projects and policies
Examples from recent advanced ABMs: Conceptual model structure
Software & hardware solutions
NYMTC, June 8, 2011 8
Topics for Discussion & Examples of Advanced Features
NYMTC, June 8, 2011 9
Feature MORPC
ARC/MTC
SACOG
SANDAG
PS RC
MAG/PAG
JT MT
BMC C MAP
SC AG
PopSyn √ √ √
Intra-HH √ √ √ √ √ √
Short/long √ √
TOD √ √ √ √ √
Cong./price √ √ √
Spec. Mark. √
Transit/sp. √ √ √
Comm. Pat. √ √
Parking √ √
Emissions √ √
ABM-DTA √ √ √
Population Synthesis
Create list of HHs with person attributes based on: Controls specified for
each zone Sample of individual
HHs (PUMS, ACS)
Majority of existing procedures (including HAJ of BPM-1): HH-level controls only Empirical algorithm
NYMTC, June 8, 2011 10
PopSyn
Input zonal values Household distributionSeed distribution
from PUMS
Population & HHs By size (1-6)
Average HH income By income group (1-3)
By size & income (6x3) By size & income (6x3)
By size income &
workers (6x3x4)
By size income workers
& children (6x3x4x4)
List of HHs for
micro-simulation
Labor force Workers % (0,1,2,3+) by HH
size & income
Children % (0,1,2,3+) by HH
size income & workers
% HH curves (1,2,3,4,5,6+) as
function of average HH size
% HH curves (1,2,3) as function
of average HH income
SANDAG PopSyn II Features
Formulated as an entropy-maximization problem
Balance person and household controls simultaneously
Applicable to both Census 2000 and ACS data
Updated household weight discretizing step
Added household allocation from TAZ to small geography
Database-driven and OOD
PopSyn
11NYMTC, June 8, 2011
New Balancing ProcedureType Controls A priori
weightsContribution coefficients
Multidimensional Matrix (CMF)
Row/column totals
Initial matrix Cell-row/columnincidence (0,1)
Table of categories (ARC)
Column totals Initial weight for category (row)
Row/column incidence (0,1)
Table of individual records (SANDAG)
Column totals Initial individual weight (row)
Row/columncoefficient (≥0)G
enera
lization
Each subsequent method includes the previous one as a particular case and guarantees the same result
Not every table of categories can be reduced to a matrix form! Not every table of individual records can be reduced to table of
categories!
12
PopSyn
NYMTC, June 8, 2011
Example of Table of HHs
HH ID HH size Person age HH initial weight
1 2 3 4+ 0-15 16-35 36-64 65+
1i 2i 3i 4i 5i 6i 7i 8i n
1n 1 1 20 2n 1 1 1 20 3n 1 1 2 20 4n 1 2 2 20 5n 1 1 3 2 20
…. … Control 100 200 250 300 400 400 650 250
13
PopSyn
NYMTC, June 8, 2011
Program Formulation
- Preserve initial HH weights as much as possible
- Meet all controls
Convex mathematical program with linear constraints Efficient method is applied to find solution:
14
PopSyn
NYMTC, June 8, 2011
SANDAG PopSyn Key Steps
Create Sample HHs
Balance HH Weights
Discretize HH Weights
Allocate HHs
Validate PopSyn
Create control targets
Create validation measures
PopSyn
15NYMTC, June 8, 2011
Control Variables
Household-level controls Household size (1,2,3,4+) Household income (5 categories) Number of workers per household (0, 1, 2, 3+) Number of children in household (0, 1+) Dwelling unit type (3 categories) Group quarter status (4 categories)
Person-level controls Age (7 categories) Gender (2 categories) Race (8 categories)
PopSyn
16NYMTC, June 8, 2011
Results – HH Characteristics
PopSyn
17NYMTC, June 8, 2011
Results – Person Characteristics
PopSyn
18NYMTC, June 8, 2011
NYMTC, June 8, 2011
Person Types in ABM
NUMBER PERSON-TYPE AGE WORK STATUS SCHOOL STATUS
1 Full-time worker 18+ Full-time None
2 Part-time worker 18+ Part-time None
3 Non-working adult 18 – 64 Unemployed None
4 Non-working senior 65+ Unemployed None
5 College student 18+ Any College +
6 Driving age student 16-17 Any Pre-college
7 Non-driving student 6 – 16 None Pre-college
8 Pre-school 0-5 None None
19
PopSyn
Intra-Household InteractionsType
Co
lum
bu
s
La
ke
Ta
ho
e
Atl
an
ta
Ba
y A
rea
Sa
n D
ieg
o
Ph
oe
nix
Je
rusa
lem
Ch
ica
go
Coordinated daily activity pattern type (sequential)
√ √
Coordinated daily activity pattern type (simultaneous)
√ √ √ √ √ √
Fully joint tours for shared non-mandatory activities
√ √ √ √ √ √ √ √
Allocated maintenance tasks √ √ √ √ √ √
Escorting children to school √ √
Car allocation √
NYMTC, June 8, 2011 20
Intra-HH
Why Intra-Household Interactions? For each worker:
80% probability of going to work
20% probability of not going to work
In 2-worker HH following IDAP:
64%=80%×80% - both workers going to work
4%=20%×20% - neither of workers going to work
32% - one of workers going to work
In 2-worker HH (observed and CDAP):
72% - both workers going to work
10% - neither of workers going to work
18% - one of workers going to work
21NYMTC, June 8, 2011
Intra-HH
Coordinated Daily Activity-travel Pattern (CDAP)
Signature feature of CT-RAMP: Trinary choice for each HH member:
(M) Mandatory pattern (N) Non-mandatory travel active pattern (H) Home (non-travel pattern)
Modeled simultaneously for all HH members taking into account coordination between them
Two significant improvements for San-Diego and Phoenix ABM: Binary sub-choice of joint non-mandatory activity
added Intermediate nesting level to account for similarity
between N and H patterns
22NYMTC, June 8, 2011
Intra-HH
CDAP for 2 Persons
23NYMTC, June 8, 2011
Intra-HH
Why Joint Travel Explicitly?
Almost 50% of tours fully or partially joint: Destinations, modes, and TOD cannot be
modeled independently
HOV/HOT lanes: Propensity to carpool and constraints are
essential
Cannot be modeled through simple mode choice
NYMTC, June 8, 2011 24
Intra-HH
25
Travel Tours by Type
0%
10%
20%
30%
40%
50%
60%
70%
Ind Fully
Joint
Joint
Out
Joint
In
Drop
off
Get
out
Pick
up
Get-
in
NY Mid Ohio
Intra-HH
NYMTC, June 8, 2011
26
Choice Alternatives
Household generation of joint tours
Travel party composition for each joint tour
No tours
1 tour
2 tours
Shopping
Eating out
Maintenance
Discretionary
Shop/shop
Shop/eat
Shop/maint
Shop/discr
Eat/eat
Eat/maint
Eat/discr
Maint/maint
Maint/discr
Adults
Children
Mixed
Discr/discr
Intra-HH
NYMTC, June 8, 2011
Person participationin each party
Yes
No
Escorting Children to School
NYMTC, June 8, 2011 27
School tour allocation by half-tours
Outbound half-tour Inbound half-tour
Ride-sharing
Pure escort
Noescort
Ride-sharing
Pure escort
Noescort
1st 2nd 3rd 1st 2nd 3rd 1st 2nd 3rd 1st 2nd 3rd
1 2 3 4 5 6 7 1 2 3 4 5 6 7
Chauffeur outbound
half-tours
Numbered half-tour alternatives
49 entire-tour
combinations
from 11 to 77
Chauffeur inbound
half-tours Available chauffeurs Available chauffeurs
Intra-HH
Importance of Workplace Choice
Cornerstone of travel demand New observed phenomena and tendencies:
More specialized occupations Growing share of work from home, flexible
schedules, & telecommuting
Advantages of ABM framework: Directly comparable to Census/ACS Unlimited segmentation (occupation, income,
gender) Disaggregate estimation & application of utility
functions
NYMTC, June 8, 2011 28
Long & short
Workplace Location Choice Utility
NYMTC, June 8, 2011 29
Occupation Person type
Residential zone
Workplace zone
/Zone size term (relevant jobs)
/Mode choice logsum
/Distance decay function
/Agglomeration & competition effects
Mode
Elemental functions
Competing locations
Long & short
Distance Decay Function
Linear combination of elemental distance (D) functions: LN(D) D0.5
D D2
D3
Great degree of flexibility in describing various non-linear effects
NYMTC, June 8, 2011 30
0
10
20
30
40
50
60
0 10 20 30 40 50 60 70
De
cay
Distance
0
0.5
1
1.5
2
2.5
0 10 20 30 40 50 60 70
De
cay
Distance
Long & short
3 Metropolitan Regions
NYMTC, June 8, 2011 31
Characteristic San Diego, CA Phoenix, AZ Tucson, AZ
MPO for which the ABM is developed
San Diego Association of Governments (SANDAG)
Maricopa Association of Governments (MAG)
Pima Association of Governments (PAG)
Population 3,095,000 4,261,000 1,035,000
HHs in the survey 3,651 3,357 1,710
Workers in the survey
4,151 3,001 1,323
Working from home
11.2% (466) 13.5% (405) 14.2% (188)
Long & short
Segmentation of Workers and Jobs by Occupation (MAG/PAG)
Workers in NHTS 2008 are classified by 5 occupation categories: Sales, marketing Clerical, administrative, retail, Production, construction, farming, transport Professional, managerial, technical Personal care or services
Jobs in each TAZ are classified by 2-digit NAICS codes (26 categories)
26 to 5 correspondence used to segment the size variables by 5 categories
NYMTC, June 8, 2011 32
Long & short
Segmentation of Distance Decay Functions
NYMTC, June 8, 2011 33
2 worker status categories: Full-time (30+hours per week) Part-time (<30 hours per week)
3 gender / household composition categories: Male Female w/child U6 Female w/o child U6
3 household income groups: Low (<$50K) Medium ($50K-$100K) High ($100K+)
Results in 2×3×3=18 segments
Long & short
Estimation of Distance Decay Functions
Baseline worker case: Male Full-time Medium HH income ($50K-$100K)
Main impacts on top of the baseline found in all 3 regions: Female gender:
With preschool child U6 W/o preschool child U6
Part-time Low income (<$50K) High income (>=$100K)
Mode choice logsum coefficient kept 0.5 across all three regions that is close to the original estimated values
NYMTC, June 8, 2011 34
Long & short
Baseline Distance Decay
NYMTC, June 8, 2011 35-12
-10
-8
-6
-4
-2
0
0 10 20 30 40 50 60 70 80 90
Uti
ls
Distance, miles
SANDAG
MAG
PAG
SANDAG jobs are closer to population compared to MAG while PAG is a smaller compact region
Long & short
Impact of Part-Time Work
NYMTC, June 8, 2011 36-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0
Uti
ls
Distance, miles
SANDAG
MAG
PAG
Part-time workers look for local jobs; the tendency is most prominent in small regions like PAG for short commuting under 10 miles (majority of cases)
Long & short
Impact of Low Income
NYMTC, June 8, 2011 37-6
-5
-4
-3
-2
-1
0
1
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0
Uti
ls
Distance, miles
SANDAG
MAG
PAG
Low-income workers look for local jobs and are less specialized in occupation; the tendency is less prominent in small regions like PAG
Long & short
Impact of High Income
NYMTC, June 8, 2011 38-0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0
Uti
ls
Distance, miles
SANDAG
MAG
PAG
High-income workers do not look for local jobs; for MAG high-income workers could not be distinguished from medium-income workers (baseline)
Long & short
Impact of Female Gender
NYMTC, June 8, 2011 39-8
-7
-6
-5
-4
-3
-2
-1
0
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0
Uti
ls
Distance, miles
SANDAG
MAG-w/o child U6
PAG-w/o child U6
MAG-w/child U6
PAG-w/child U6
There is still a gender bias; females, especially with small children tend to avoid long-distance commuting; w/o children the bias is less prominent, especially in a small region like PAG
Long & short
Composition of All Impacts (MAG)
NYMTC, June 8, 2011 40
Long & short
Validation, SANDAG, 8×8 Major Statistical Areas
NYMTC, June 8, 2011 41
0
50,000
100,000
150,000
200,000
250,000
0 50000 100000 150000 200000 250000
Esti
mat
ed
Wo
rke
r Fl
ow
s (N
orm
aliz
ed
)
CTPP Worker Flows
Normalized Estimated
Linear (trend)
No K-factors needed!
Long & short
Summary of Main Factors Segmentation by occupation to connect right workers by
place of residence to right jobs Commuting distance has a complex non-linear effect on
workplace choice differentiated by person type: Constrained time budget results in cut-off thresholds (40-60
min) Minimal commuting time is acceptable and usable resulting
in a low-sensitivity region (0-30 min)
Incorporation of these non-linear effects in mode choice logsum instead of distance-based terms: Theoretically appealing Practically difficult to achieve: mode choice and destination
choice are subject to different considerations, time scales, and constraints
NYMTC, June 8, 2011 42
Long & short
Advanced TOD Choice Models
Consider tours and activity durations, not just trips
Operate at fine level of temporal resolution: 30 min or less in discrete implementation
First continuous implementations
Generate consistent individual daily schedule w/o gaps or overlaps
NYMTC, June 8, 2011 43
TOD
NYMTC, June 8, 2011 44
Tour TOD Choice
5 23
Work tour to schedule
TOD
NYMTC, June 8, 2011 45
Tour TOD Choice
5 23
Work tour to schedule
Considerations for departure time:
•Office hours (7-10)
•Avoid congestion (10+)
•Give ride to child (7)
TOD
NYMTC, June 8, 2011 46
Tour TOD Choice
5 23
Work tour
10
TOD
NYMTC, June 8, 2011 47
Tour TOD Choice
5 23
Work tour
10
Considerations for arrival time:
•Office hours (<=20)
•Avoid congestion (<16)
•Tennis before dark (<17)
TOD
NYMTC, June 8, 2011 48
Tour TOD Choice
5 23
Work tour
10 15
Considerations for duration:
•8 work hours
•Finish presentation for NYMTC
TOD
NYMTC, June 8, 2011 49
Tour TOD Choice
5 23
Work tour
9 19
TOD
TOD Choice Dimensions Formal (820):
40 departure half-hours (5AM-24PM) by
40 arrival half-hours (departure-24PM) leads to
40×41/2=820 feasible combinations
Real & meaningful (120): 40 departure half-
hours and 40 arrival half-hours
and 40 durations
50
Departure Alternatives
Arrival Alternatives
1 2 3 4 . . n
1
2
3
4
.
.
n
TOD
NYMTC, June 8, 2011
NYMTC, June 8, 2011 51
Sequential Processing of Tours
5 23
1-Work
2-Discretionary joint
3-Shopping individual
TOD
NYMTC, June 8, 2011 52
Sequential Processing of Tours
5 23
1-Work
7-17
2-Discretionary joint
3-Shopping individual
TOD
NYMTC, June 8, 2011 53
Sequential Processing of Tours
5 23
1-Work
7-17
2-Discret
20-23
3-Shopping individual
TOD
NYMTC, June 8, 2011 54
Sequential Processing of Tours
5 23
1-Work
7-17
2-Discret
20-23
3-Sh
18-19
TOD
Calibration Results –SANDAG/Work
55NYMTC, June 8, 2011
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
Befo
re 5
am
5:0
0 a
m t
o 5
:30 a
m
5:3
0 a
m t
o 6
:00 a
m
6:0
0 a
m t
o 6
:30 a
m
6:3
0 a
m t
o 7
:00 a
m
7:0
0 a
m t
o 7
:30 a
m
7:3
0 a
m t
o 8
:00 a
m
8:0
0 a
m t
o 8
:30 a
m
8:3
0 a
m t
o 9
:00 a
m
9:0
0 a
m t
o 9
:30 a
m
9:3
0 a
m t
o 1
0:0
0 a
m
10:0
0 a
m t
o 1
0:3
0 a
m
10:3
0 a
m t
o 1
1:0
0 a
m
11:0
0 a
m t
o 1
1:3
0 a
m
11:3
0 a
m t
o 1
2:0
0 p
m
12:0
0 p
m t
o 1
2:3
0 p
m
12:3
0 p
m t
o 1
:00 p
m
1:0
0 p
m t
o 1
:30 p
m
1:3
0 p
m t
o 2
:00 p
m
2:0
0 p
m t
o 2
:30 p
m
2:3
0 p
m t
o 3
:00 p
m
3:0
0 p
m t
o 3
:30 p
m
3:3
0 p
m t
o 4
:00 p
m
4:0
0 p
m t
o 4
:30 p
m
4:3
0 p
m t
o 5
:00 p
m
5:0
0 p
m t
o 5
:30 p
m
5:3
0 p
m t
o 6
:00 p
m
6:0
0 p
m t
o 6
:30 p
m
6:3
0 p
m t
o 7
:00 p
m
7:0
0 p
m t
o 7
:30 p
m
7:3
0 p
m t
o 8
:00 p
m
8:0
0 p
m t
o 8
:30 p
m
8:3
0 p
m t
o 9
:00 p
m
9:0
0 p
m t
o 9
:30 p
m
9:3
0 p
m t
o 1
0:0
0 p
m
10:0
0 p
m t
o 1
0:3
0 p
m
10:3
0 p
m t
o 1
1:0
0 p
m
11:0
0 p
m t
o 1
1:3
0 p
m
11:3
0 p
m t
o 1
2:0
0 a
m
Aft
er
12:0
0 a
m
Work Departure Observed Work Departure Estimated Work Arrival Observed Work Arrival Estimated
TOD
Calibration Results –SANDAG/Work
56NYMTC, June 8, 2011
0%
2%
4%
6%
8%
10%
12%
14%
0 h
ours
0.5
hours
1 h
ours
1.5
hours
2 h
ours
2.5
hours
3 h
ours
3.5
hours
4 h
ours
4.5
hours
5 h
ours
5.5
hours
6 h
ours
6.5
hours
7 h
ours
7.5
hours
8 h
ours
8.5
hours
9 h
ours
9.5
hours
10 h
ours
10.5
hours
11 h
ours
11.5
hours
12 h
ours
12.5
hours
13 h
ours
13.5
hours
14 h
ours
14.5
hours
15 h
ours
15.5
hours
16 h
ours
16.5
hours
17 h
ours
17.5
hours
18 h
ours
18.5
hours
19 h
ours
19.5
hours
Work Duration Observed Work Duration Estimated
TOD
Basic Generalized Cost Function (Starting Point of SHRP 2 C04)
U=b×Time+c×Cost b = travel time coefficient
c = travel cost coefficient
VOT = b/c (constant)
99% of research and 100% of models in practice use this function
This function is simplistic and masks many important effects of congestion and pricing
57
Congestion & Pricing
NYMTC, June 8, 2011
VOT Growth with Journey Lengths
Cost damping: Poor perception of car operating cost vs. parking and tolls Relative rather than absolute perception of cost Cheaper housing and higher disposable income for long-
distance commuters Trip frequency inversely proportional to trip length (for non-
work travel) Higher car occupancy for longer trips (if car occupancy is not
accounted)
Time valuing: Risk aversion (if reliability is not accounted ) Unfamiliarity with distant locations Time budget constraints
58
Congestion & Pricing
NYMTC, June 8, 2011
Non-Linear Distance Effects (NY)
U=b×Time+c×Cost
U=(b1+b2×Dist+b3×Dist2+…)×Time+c×Cost
59
Distance
VOT
30 miles
Congestion & Pricing
NYMTC, June 8, 2011
VOT Drop for Long-Distance Commuters
Self-selection of low-VOT commuters by residential choice
Long commuting time used productively (laptops, cell phones)
Restructured (simplified) daily activity pattern: Compressed work week with no other out-of-
home activities on regular workday
Compressed shopping and discretionary activities on (extended) weekends
60
Congestion & Pricing
NYMTC, June 8, 2011
Perceived Time by Congestion Levels
U=b×Time+c×Cost
U=b1×FFTime+b2×Delay+c×Cost
b2 / b1 ≈ 1.5-2.0
Every minute spend in congestion conditions is perceived as 1.5-2.0 min of free driving!
Proxy for travel time reliability:
Loses significance if reliability is incorporated directly
Useful for simple models that cannot incorporate reliability directly
61
Congestion & Pricing
NYMTC, June 8, 2011
Impact of Income on Sensitivity to Cost (NY)
U=b×Time+c×Cost
U=b×Time+c×(Cost / Ince)
e ≈ 0.5-0.7
VOT grows with income (constant elasticity)
Commuting VOT range:
62
Household Income VOT
$12,500 $5/hour
$25,000 $8/hour
$50,000 $15/hour
$75,000 $22/hour
$100,000 $27/hour
$150,000 $39/hour
$200,000 $46/hour
Congestion & Pricing
NYMTC, June 8, 2011
Impact of Car Occupancy (NY) U=b×Time+c×Cost
U=b×Time+c×(Cost / Occf) f ≈ 0.6-0.8
VOT grows with occupancy but not linearly: Less cost sharing for intra-household carpools
Almost proportional cost sharing for inter-household carpools
Typical cost sharing: SOV=1.00
HOV2=0.57
HOV3=0.41
63
Congestion & Pricing
NYMTC, June 8, 2011
Cost Sharing Parameter for HOV
64
Intra-household
Inter-household
With
child
ren
Adults
only
1.0
0.0
Congestion & Pricing
NYMTC, June 8, 2011
Combined Income-Occupancy Effects
Captured by: Constants in mode choice framework Explicit modeling of joint travel in advanced ABMs
Low-income workers have more opportunities to form inter-household carpools: Fixed schedules Residential clusters Job clusters
Mitigates equity concerns regarding pricing: Cost is shared Low-income workers can switch from HOV or transit High-income workers can only switch to transit
65
Congestion & Pricing
NYMTC, June 8, 2011
Distributed Value of Time (SFCTA)
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
$- $5 $10 $15 $20 $25 $30
Value of T ime ($/Hour)
Pro
ba
bil
ity
De
ns
ity
Income $0-30kIncome $30-60k
Income $60-100kIncome $100k+
Congestion & Pricing
66NYMTC, June 8, 2011
NYMTC, June 8, 2011
Activity Types
67
TYPE PURPOSE DESCRIPTION CLASSIFICATION ELIGIBILITY
1 Work Working at regular workplace or work-related activities outside the home.
Mandatory Workers and students
2 University College + Mandatory Age 18+
3 High School Grades 9-12 Mandatory Age 14-17
4 Grade School Grades K-8 Mandatory Age 5-13
5 Escorting Pick-up/drop-off passengers (auto trips only).
Maintenance Age 16+
6 Shopping Shopping away from home. Maintenance 5+ (if joint travel, all persons)
7 Other Maintenance Personal business/services, and medical appointments.
Maintenance 5+ (if joint travel, all persons)
8 Social/Recreational Recreation, visiting friends/family.
Discretionary 5+ (if joint travel, all persons)
9 Eat Out Eating outside of home. Discretionary 5+ (if joint travel, all persons)
10 Other Discretionary Volunteer work, religious activities.
Discretionary 5+ (if joint travel, all persons)
Markets
Extended Population Synthesis (MAG)
NYMTC, June 8, 2011 68
Markets
Pop
ulat
ion
Syn
thes
isU
sual
wor
kpla
ceU
sual
Sch
ool
1.1. Core
HH
population
and GQ
2.1.1-2.1.2
2.1.3-2.1.4
Major universities
Permanent Seasonal
Students in dorms
(1-person)
Students in rent apt (1-person)
Stock for rent
Students control
1.2.1. Fixed on-campus
1.2.2. Off-campus choice
Enr
ollm
ent
Em
ploy
men
t
1.3. Seasonal residents
1.4.1. Visitors for business
(1-person)
1.4.2. Visitors for recreation
1.5. Transient population
Hotels occupancy
Duration of stay
Duration of stay
2.1.2. Agriculture & construction
jobs
2.1.2. Business meeting location
Conventional
components
New endogenous components
New exogenous components
Integration of Special Events (MAG)
NYMTC, June 8, 2011 69
Markets
Treatment of Space
Level of spatial resolution: TAZ (3,000-4,000)
MGRA (20,000-30,000)
Parcel (1,000,000)
Calculation of LOS: Predetermined Origin and Destination
catchment areas
On-fly path building
NYMTC, June 8, 2011 70
Transit
Fine-Grain LOS (1=Pre-fixed)
NYMTC, June 8, 2011 71
Origin 2 TAZ/TAP
Destination 1 TAZ/TAP
Origin 1 TAZ/TAP
Origin 3 TAZ/TAP
Destination 2 TAZ/TAP
Destination 3 TAZ/TAP
2
3
1
5
6
4
8
9
7
2
3
1
5
6
4
8
9
7
Access EgressMain In-Vehicle
Transit
Fine-Grain LOS (2=on Fly)
NYMTC, June 8, 2011 72
Origin 2 Stop
Destination 1 Stop
Origin 1 Stop
Origin 3 Stop
Destination 2 Stop
Destination 3 Stop
2
3
1
5
6
4
8
9
7
2
3
1
5
6
4
8
9
7
Access EgressStop-to-Stop LOS
Transit
Transit Tour Modes
Main transit modes classification: Bus (local & express)
Subway, LRT, ferry (w/bus)
Commuter rail (w/subway & bus)
Access/egress modes classification: Walk/bike
K&R (including taxi)
P&R
NYMTC, June 8, 2011 73
Transit
Transit Nest StructuresTransit
Bus
Walk P&R K&R
Subway, LRT, ferry
Walk P&R K&R
Commuter rail
Walk P&R K&R
NYMTC, June 8, 2011 74
Transit
Walk
Bus Subway Rail
P&R
Bus Subway Rail
K&R
Bus Subway Rail
Transit
Advanced Cross-Nested Structure
NYMTC, June 8, 2011 75
Walk to bus
P&R bus
K&R bus
Walk to subway
P&R subway
K&R subway
Walk to rail
P&R rail
K&R rail
Bus Subway Rail Walk P&R K&R
Competition between main modes
Competition between access modes
Transit
Transit
Transit Crowding Function
NYMTC, June 8, 2011 76
Crowding Factor
Voltr
1.00
0 Seat Cap
Fcap
Fseat
MaxCon
Transit
Activity Analysis – Why Essential?
Substitution between in-home and out-of-home activities: Telecommuting
Teleshopping
Compressed work schedules
Impact of one activity on other activities through time-space constraints Interdependence between work and non-work trip rates
Interdependence between activity & travel of household members
77NYMTC, June 8, 2011
Commuting
General Framework for Workplace Choice
NYMTC, June 8, 2011 78
Worker characteristics:
• Person (age, occupation, gender, education, etc)
• HH (income, composition, age of children)
• Residential location (accessibility to relevant jobs)
Work at home
permanently
Usual workplace
out of home
TAZ 1:Jobs
TAZ 2:Jobs
TAZ N:Jobs
…
Individual accessibility
Commuting
Workplace Type Choice Utility
NYMTC, June 8, 2011 79
Work out of home:
Work at home:
Occupation Person type
Residential zone
Workplace zone
Workplace zone choice utility
Accessibility to jobs
Person & HH attributes
Commuting
Workplace Type Choice –Work from Home (MAG/PAG)
NYMTC, June 8, 2011 80
Variable Coefficient t-stat
Constants General -0.851 -2.46Tucson -0.034 -0.33
Status Full Time Worker -1.178 -11.04Gender Female -0.346 -3.43
Household composition
Female Worker with Preschool Child Child in the HH 0.382 1.68
Non-Working Adults in the HH -0.192 -1.54
Occupation Sales or marketing 0.765 5.89
Age Group
Age <= 35 years -0.230 -1.31
35 years to 44 years (reference)
45 years to 54 years 0.332 2.3455 years to 64 years 0.348 2.37Age 65 years or older 0.432 2.38
Household Income group
$49,999 or Less -0.090 -0.63
$50,000 to $74,999 (reference)
$75,000 to $99,999 0.160 1.07$100,000 or more 0.267 1.95
Education Level
Less than High School Educated -0.398 -0.95
High School completed (reference)Bachelor's or Some College degree holder 0.295 2.28
Master's or higher degree holder 0.300 1.89
Accessibility Accessibility to Employment Locations by Job Category (Logged)
-0.069 -2.22
Model stats
Number of Observations 4,324
Likelihood with Constants only -1728.4776
Final likelihood -1601.4239
Rho-Squared (0): 0.4657
Rho-Squared (constant): 0.0735
Commuting
Predicting Future for Working from Home & Telecommuting
Rapidly growing %: Work from home Full or partial telecommuting Compressed & flexible work schedules
Result of: Communication technology Structural shifts in occupation and industries
One of the biggest unknowns: Saturation or trends will hold?
Significant impacts on congestion levels (reduction) and VMT (mixed): Effective policy variable Sensitivity tests possible with model that has this component as
policy lever
NYMTC, June 8, 2011 81
Commuting
Taxonomy of Work Arrangements
Arrangement Normal Alternative
Job type Full time (30+ hours) Part-time (≤29 hours)
Number of jobs 1 2+
Usual workplace location
Out of home -permanent
At home
Out of home - variable
Commuting frequency 5 days a week 1-4 days a week (compressed)
6-7 days a week (extended)
Telecommuting frequency
Low (less than once a week)
High (once a week or greater)
Schedule flexibility No / little Yes / significant
Usual schedule AM / PM Second shift, otherNYMTC, June 8, 2011 82
Commuting
Work Arrangements:Sequence of Choices
NYMTC, June 8, 2011 83
Person & household chacteristics
Occupation
Job type:
1=full-time, 2=part-time
Schedule flexibility:
1=no, 2=some, 3=free
Workplace
Home Outside home
Workplace Location
1/7 2/7 3/7 4/7 5/7 6/7 7/7
General commuting frequency
(days at work)
Telecommuting frequency
6=No5=Once a
year
4=Few
times a
year
3=once a
month or
more
2=once a
week or
more
1=almost
every day
Number of jobs: 1, 2+
1. Main work
arrangements
2. Commuting
frequency &
flexibility
Commuting
NYMTC, June 8, 2011 84
Aggregation Bias Example: Parking Cost
Auto Share
0%
20%
40%
60%
80%
100%
0 4 8 12 16 20
Parking Cost, $
Before:50% - no charge
50% - $16
Average - $8
After:50% - $4
50% - $20
Average - $12
Shift:From 70% to 30% by
average
From 52.5% to 47.5% structural
Parking
NYMTC, June 8, 2011 85
Parking Demand
Travel-related attributes (endogenous to ABM): Tour/trip destination
Arrival time
Planned activity duration (parking space occupied)
Person-related attributes: Person socio-economic characteristics
Travel party size
Willingness to pay versus parking search and/or walk associated with remote parking
Eligibility for free / subsidized parking
Parking
NYMTC, June 8, 2011 86
Parking Supply in each Zone
Free parking:
Capacity
Paid parking:
Capacity
Rate:
Daily (for long parking)
Hourly (for short parking)
Parking
NYMTC, June 8, 2011 87
Parking Equilibrium ModelIndividual Demand for Parking
Parking supply
Demand-supply equilibrium
Person characteristics and willingness to pay
Free parking eligibility
Travel purpose, destination, and arrival time
Duration of activity
Travel party
Free and paid parking capacity
Parking rate (daily/hourly)
Location and distance from the destination
Parking location choice
Parking search time
Parking occupancy and residual availability
Impact on
destination choice
and schedule
corrections
Impact of parking
scarcity on rate
(price mechanism)
Parking
Examples of Estimated Models
Included in Columbus, OH (MORPC) and San-Diego (SANDAG) ABMs
Based on Household Travel Survey and complementary Parking Survey in downtown
Choice models applied for each tour: Parking provision for downtown destinations (free
on-site, off-site reimbursement, no provision)
Reimbursement amount
Parking lot location for downtown destinations
NYMTC, June 8, 2011 88
Parking
Parking Location Choice
In most applied models, parking lot is assumed in the destination zone
In reality, parking constraints and congestion may lead to remote parking
NYMTC, June 8, 2011 89
Origin zone
Destination zone
Parking zone
Intended auto trip
Actual auto trip Walk
Parking
Parking Locations (San Diego)
NYMTC, June 8, 2011 90
Parking
Parking Location Choice
-6 -4 -2 0 2
Parking cost, cents
Walk distance, miles
Log of parking capacity
Auto time, min
Same area type as destination
CBD
Free, non-work
Free, work
Paid, non-work
Paid, work
NYMTC, June 8, 2011 91
Parking
Car Type Choice
NYMTC, June 8, 2011 92
Car type
Car Type Choice
NYMTC, June 8, 2011 93
Car type
ABM-DTA Integration Dilemma
NYMTC, June 8, 2011 94
ABM-DTA
Microsimulation ABM
Microsimulation DTA
List of
individual
trips
Individual
trajectories
for the
current list of
trips
LOS for
the other
potential
trips?
New Approach (SHRP 2 L04)
NYMTC, June 8, 2011 95
Microsimulation ABM
Microsimulation DTA
List of
individual
trips
Individual
trajectories
for the
current list of
trips
Consolidation of individual
schedules (inner loop for
departure / arrival time
corrections)
Sample of alternative origins,
destinations, and departure times
Individual
trajectories
for potential
trips
ABM-DTA
Temporal equilibrium to achieve individual schedule consistency
Individual Schedule Consistency
NYMTC, June 8, 2011 96
0 24
Activity i=0 Activity i=1 Activity i=2
Trip i=1 Trip i=2 Trip i=3
Activity i=3
Departure
Arrival
Duration
Travel
id
iT
i
i
Schedule
i
ABM-DTA
Conclusions
Tremendous progress in ABM since 2002 when BPM 1.0 was created
Combination of most advanced features of existing ABMs is good starting point to specify BPM 2.0
Integration of advanced features and addressing unique NY conditions in computationally efficient way is still challenge
NYMTC, June 8, 2011 97