1. Evan Roux, Theuns Lamprecht Tolplan (PTY) Ltd
APPLICATION OF STATED PREFERENCE CHOICE MODELS IN THE CITY OF JOHANNESBURG TRANSPORT DEMAND MODEL
Evan Roux, Theuns Lamprecht
Tolplan (PTY) Ltd

2. Content: Background COJ Transport demand model design SP surveys
Mode choice design
Application of choice model in demand model
Model validation
Model outputs
Conclusions

3. Background Study originated from: Acknowledgments:
City of Johannesburg Integrated Transport Network 2015
Required an updated transport demand model
Significant focus on BRT services
Acknowledgments:
City of Johannesburg –Transport Department
Prof Christo Venter – University of Pretoria (SP Surveys)

4. City of Johannesburg Largest city in South Africa
Situated in Gauteng province
Covers an area of 1 645km2
Population of 4.4 million (2011)

5. COJ Transport Demand Model Design:
Household travel survey & other
Productions & Attractions and Utilities
Demand Matrix
Private & Public Demand Matrices
Trip Generation
Trip Distribution
Mode Split
Route Assignment
Productions & Attractions
Demand Matrix
Private & Public Demand Matrices
Impedance (Skims)

6. COJ Transport Demand Model Design

7. SP surveys: Fieldwork: Entire survey on tablet computers
Pilots 1 & 2: June 2014
Main survey: July-August 2014
Total sample of 1208 respondents
30 fieldworkers mostly from communities
Entire survey on tablet computers
“Pivot design” using respondent’s actual trip experience as starting point
Attempted to understand & model captivity
Model estimation
Mixed logit
Included RP and SP in same model to improve validity

8. Survey Design Attribute Levels Mode constant (current mode)
Car, Gautrain, Taxi, Bus, BRT, Metrorail
Number of transfers (PT only)
No transfers; 1 transfer
Travel cost
-20%; current; +20%
In-vehicle travel time
Walk time to PT
-50%; current; +50%
Wait time for PT

9. Example: SP Experiment

10. Mode Choice Model Design

11. Mode Choice Model Design
Estimated coefficients
AVAIL CARCAP & CHOOSERS
PT CAPTIVES
COEFFICIENT
VARIABLE
UNIT LI MI HI
BUS
BRT
0.0000
CAR
1.8990 --
GAUTR
2.5500
TAXI
TRAIN
COST
Rands
IN-VEH TIME
Minutes
WALK TIME START OF TRIP
WAITING TIME
SEAT AVAILABLE ON BRT*
1=Yes
0.0345
NO OF TRANSFERS
Number

12. Logit Mode split model – E.g. Low income PUT captive users
𝑃𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑢𝑠𝑖𝑛𝑔 𝑚𝑜𝑑𝑒 𝑎= 𝑒 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑚𝑜𝑑𝑒 𝑎 𝑒 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑚𝑜𝑑𝑒 𝑎 + 𝑒 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑚𝑜𝑑𝑒 𝑏 +𝑒 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑚𝑜𝑑𝑒 𝑐
𝑈𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑚𝑜𝑑𝑒=𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 𝑠𝑝𝑒𝑠𝑖𝑓𝑖𝑐 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡+ 𝑇𝑟𝑎𝑣𝑒𝑙 𝐶𝑜𝑠𝑡𝑠 𝑟𝑎𝑛𝑑𝑠 × − 𝐼𝑛 𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝑡𝑖𝑚𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑊𝑎𝑙𝑘 𝑡𝑖𝑚𝑒 𝑡𝑜 𝑠𝑡𝑎𝑟𝑡 𝑜𝑓 𝑡𝑟𝑖𝑝 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑊𝑎𝑖𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑆𝑒𝑎𝑡 𝑎𝑣𝑖𝑎𝑙𝑎𝑏𝑙𝑒 𝑜𝑛 𝐵𝑅𝑇 ×0.345+𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑠×(− =𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 𝑠𝑝𝑒𝑠𝑖𝑓𝑖𝑐 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡+ 𝑇𝑟𝑎𝑣𝑒𝑙 𝐶𝑜𝑠𝑡𝑠 𝑟𝑎𝑛𝑑𝑠 × − 𝐼𝑛 𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝑡𝑖𝑚𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑊𝑎𝑙𝑘 𝑡𝑖𝑚𝑒 𝑡𝑜 𝑠𝑡𝑎𝑟𝑡 𝑜𝑓 𝑡𝑟𝑖𝑝 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑊𝑎𝑖𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑚𝑜𝑑𝑒=𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 𝑠𝑝𝑒𝑠𝑖𝑓𝑖𝑐 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡+ 𝑇𝑟𝑎𝑣𝑒𝑙 𝐶𝑜𝑠𝑡𝑠 𝑟𝑎𝑛𝑑𝑠 × − 𝐼𝑛 𝑣𝑒ℎ𝑖𝑐𝑙𝑒 𝑡𝑖𝑚𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑊𝑎𝑙𝑘 𝑡𝑖𝑚𝑒 𝑡𝑜 𝑠𝑡𝑎𝑟𝑡 𝑜𝑓 𝑡𝑟𝑖𝑝 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑊𝑎𝑖𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 × − 𝑆𝑒𝑎𝑡 𝑎𝑣𝑖𝑎𝑙𝑎𝑏𝑙𝑒 𝑜𝑛 𝐵𝑅𝑇 ×0.345+𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑠×(−0.0586
𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑉𝐼𝑆𝑈𝑀 𝐷𝑒𝑚𝑎𝑛𝑑 𝑝𝑟𝑜𝑐𝑒𝑑𝑢𝑟𝑒

13. Alternative specific constants
Order of attractiveness of modes (based on constants from primary mode choice model), at identical cost and time attributes. Reflects perceptions of qualitative difference between modes as currently experienced.

14. Application of choice model in demand model
Trip Generation Model
Trip Distribution Model
Low income persons
Home base work trips
Home based education trips
Home based shopping trips
Home based other trips
Non home based trips
Medium income persons
High income persons
Stratified demand

15. Application of choice model in demand model

16. Application of choice model in demand model
Average proportion of captives applied in the transport model
Income level
Car captive
(Lifestyle)
PT captive
Choosers
(Availability)
Total
Low
9.5%
64.6%
9.9%
16.0%
100.0%
Medium
12.6%
56.4%
10.5%
20.6%
High
21.4%
37.8%
11.2%
29.6%

17. Application of choice model in demand model
Sum 5 low income distribution matrices
Matrix value x % PUT captives for origin zone
PUT captive mode split model
Bus
Taxi
BRT
Rail
Gautrain
Matrix value x % Choosers for origin zone &
Matrix value x % Car alternative captives for origin zone
Choosers and Car alternative mode split model
Car
Matrix value x % Car leisure captives
Private vehicle matrix
Assignment Matrices

18. Model validation Comparison between modelled and HTS modal shares Mode
CoJ Model Mode Split 2014
HTS Mode Split 2013
Private Car
45.2%
BRT
3.0%
1.2%
Bus
14.2%
8.0%
Gautrain
0.6%
0.1%
Taxi
29.2%
38.3%
Metrorail
7.8%
4.8%
Other 0%
2.4%
TOTAL
100%
Except for Gautrain, limited public transport census data on line level

19. Outputs used in the 2015 COJ ITN
PUT AM Peak Hour (1 hour)
- Total Network Assignment

20. Outputs of hypothetical scenarios
Modelled mode split results of hypothetical scenarios
Mode
2025 AM BRT Network (Reference)
Increase BRT headway (3min to 10min)
Increase existing GFIP tolls (100% increase)
Increase in fuel price (20% increase)
Improved Metrorail Perception (=Gautrain)
Private Car
49.0%
49.5%
48.5%
46.3%
48.0%
BRT
14.1%
11.9%
14.2%
14.8%
13.5%
Bus
9.3%
9.8%
9.4%
8.8%
Gautrain
1.6%
1.7%
Taxi
20.3%
21.1%
20.5%
21.3%
19.6%
Metrorail
5.8%
5.9%
6.1%
8.5%
Total
100.0%

21. Conclusions
VISUM already includes the functionality to use various types of mode choice models, but it is rarely used in SA.
COJ model demonstrated the power of the interaction and cumulative effect of typical transport model variables (route choice, speeds, travel time, utilisation, capacity, etc.) and mode choice variables (time, cost, transfers, mode preferences, etc).
COJ model still based on “main mode” demand matrices – scope to progress to multiple mode choices along a route based on perceived journey cost.
Limited PUT census data on lines, stations and stops to property calibrate and validate the public transport model results on a mode level.