Camila Balbontin is a Postgraduate Research Fellow at the Institute of Transport and Logistics Studies (ITLS) of University of Sydney. In February 2018, she completed her PhD under the supervision of Professor David Hensher where she focused on integrating decision heuristics and behavioural refinements into travel choice models. She was awarded the ITLS prize for Research Excellence in Transport or Logistics 2017. Camila also holds a bachelor degree in the field of Civil Engineering with a diploma in Industrial Engineering and in Transportation and Logistics from Pontificia Universidad Católica de Chile. She did her MSc degree at the same university under the supervision of Professor Juan de Dios Ortúzar. Her MSc thesis estimated the valuation of households and neighbourhood attributes in the centre of Santiago.
As a Postgraduate Research Fellow, her main focus is choice modelling and travel behaviour. She is currently working on projects related to the BRT Centre of Excellence, business location decisions, hybrid modelling, value uplift, among others.
Working Paper - http://sydney.edu.au/business/itls/research/publications/working_papers
Coefficient of Thermal Expansion and their Importance.pptx
Camila Balbontin - Do preferences for BRT and LRT change as a voter, citizen, tax payer, or self-interested resident?
1. The University of Sydney Page 1
Do preferences for BRT and
LRT change as a voter,
citizen, tax payer, or self-
interested resident?
Camila Balbontin
David A. Hensher
Chihn Q. Ho
Corinne Mulley
June, 2018
BRT Centre of
Excellence webinar
2. The University of Sydney Page 2
ACKNOWLEDGEMENTS
This paper contributes to the research program of the Volvo Research and
Education Foundation Bus Rapid Transit Centre of Excellence (BRT+). We
acknowledge the Foundation for funding support. The support of Theo Yeche
and Patricia Aranda in translating the survey instrument into French is greatly
appreciated. We also thank Rosario Macario (IST, Portugal), Anson Stewart
and Chris Zegras (Transportation and Urban Planning, MIT) for their
contributions in facilitating access to survey participants.
3. The University of Sydney Page 3
OVERVIEW
I. Introduction
II. Choice Experiment
III. Samples
IV. Model Formulation
V. Results
VI. Willingness to Pay
VII.Simulated Scenarios
VIII.Key Findings
4. The University of Sydney Page 4
INTRODUCTION
– Bus rapid transit (BRT) has often been passed over in favour of light
rail transit (LRT) in many geographical settings in developed
economies despite having much appeal in delivering high quality
services in a cost effective manner.
– Well known emotional attachment to rail solutions and image
perception
Australia France Portugal U.K. U.S.A.
5. The University of Sydney Page 5
INTRODUCTION
The focus of this paper is on whether there are significant differences
or similarities in the key behavioural outputs associated with five
measures of preference revelation when comparing a BRT system with
an LRT system in five countries.
Which one would benefit your metropolitan area better?
Which one do you personally prefer?
Which investment is better value for tax payers’ money?
If you were voting now, which one would you vote for?
Which investment would improve the liveability of the metropolitan
area more?
6. The University of Sydney Page 6
CHOICE EXPERIMENT
– 19 cities in 5 countries.
– Design with a BRT and a LRT
alternative, both with fixed route length.
– Each respondent answers two choice
tasks.
– At the end, individuals were asked for
their:
– Socio-demographics and their
experience on public transport
(number of times used in the last
month).
– Which attributes they had attended
and not attended to (Stated Attribute-
Country City Sample Percentage
Australia
Sydney 271 7%
Melbourne 241 6%
Canberra 100 3%
Brisbane 201 5%
All other capital cities
205 6%
US
Boston 181 5%
Los Angeles-Long
Beach
300 8%
Seattle-Bellevue-
Everett
100 3%
Minneapolis – St Paul 100 3%
Dallas – Fort Worth 150 4%
Philadelphia 170 5%
Portugal Lisbon 425 14%
U.K.
Birmingham 274 12%
Newcastle 153 6%
France
Lyon 128 8%
Toulouse 137 6%
TOTAL 3136
7. The University of Sydney Page 7
– Illustrative choice screen
of the survey
8. The University of Sydney Page 8
SAMPLES
Location Australia U.S. Portugal France U.K.
Socioeconomic Profile
Age
mean (std dev)
44.32
(14.9)
44.42
(16.11)
40.75
(11.45)
42.83
(14.41)
47.52
(15.35)
Gender female (%) 54% 61% 52% 53% 50%
Personal income in 1000 AUD$
mean (std dev)
$61.24
(43.50)
$80.57
(60.79)
$30.36
(26.11)
$50.63
(35.26)
$39.12
(31.53)
Trip profile
% of people that used public transport in
the last month
66% 49% 74% 69% 75%
Number of times using Bus/BRT in the
last month
5.85
(9.13)
7.46
(9.63)
11.46
(14.26)
7.13
(10.19)
7.27
(9.45)
Number of times using Light Rail/Tram
in the last month
1.73
(4.35)
2.14
(4.33)
2.48
(5.69)
3.91
(6.23)
0.53
(1.76)
Number of times using Train/Metro in
the last month
6.49
(9.92)
4.29
(7.89)
13.94
(13.87)
9.27
(11.58)
3.55
(6.57)
9. The University of Sydney Page 9
SAMPLES
How many respondents change their response
across the five measures of preference
revelation?
Australia
29%
France
32%
Portugal
27%
U.K.
33%
U.S.A.
33%
10. The University of Sydney Page 10
PREFERENCE MODEL ESTIMATION
Hypotheses:
– There are significant differences in the drivers and WTP estimates
when considering different perspectives
– Respondents’ actual experience has a significant influence over
preferences
– The influence of actual experience is different in the investment
characteristics, system characteristics and in the cost attributes
Compare the results of the models considering different
perspectives
Compare models with and without actual
experience
Compare a model that considers a common influence of
experience with a model considering specific influences overt
subsets of attributes
11. The University of Sydney Page 11
PREFERENCE MODEL ESTIMATION
– Actual experience is included as conditioning the utility function:
– The frequency of use per mode in the last month
– As a dummy variable indicating that the respondent used public
transport in the last month and has BRT and LRT available in
their city
To test our hypotheses, we estimate three models:
1) MNL, simple multinomial logit model
2) HMNL0, Heteroscedastic MNL (HMNL), conditioned by actual
experience
3) HMNL conditioned by actual experience specific to subset of the
attributes
12. The University of Sydney Page 12
PREFERENCE MODEL ESTIMATION
1 1 1
1
1, 1, 1
, , , ,
inv sys
scst bus metro train
MNL
BRT BRT x inv BRT x sys BRT Z
x cst BRT fr BRT bus freq BRT metro freq BRT train
U ASC x x Z
x freq freq freq
1 1 11, 1, ,
_ _
inv sys scst
MNL
LRT LRT x inv LRT x sys LRT x cst LRT
usePT BRTLRT usePT BRTLRT
U ASC x x x
dummy
Simple MNL
Investment
characteristi
cs
System
characteristi
cs
Cost
characteristi
cs
Socio-
demographics
Frequency of bus,
metro and train use in
the last month
Dummy that is one if the
respondent used public
transport in the last month and
he/she has BRT and LRT
available in their city
13. The University of Sydney Page 13
PREFERENCE MODEL ESTIMATION
HMNL0
1 1 1 1
0
, , ,
1, 1, 1 ,
(1 )
... ... ...
bus metro train
inv sys scst
HMNL
BRT fr BRT bus freq BRT metro freq BRT train
BRT x inv BRT x sys BRT Z x cst BRT
U freq freq freq
ASC x x Z x
1 1 1
0
_ _
1, 1, ,
(1
)
... ...
bus metro train
inv sys scst
HMNL
LRT freq bus freq metro freq train
usePT BRTLRT usePT BRTLRT
LRT x inv LRT x sys LRT x cst LRT
U freq freq freq
dummy
ASC x x x
Frequencies of
modes use
condition the utility
function
Same with the
dummy variable
representing PT
use with BRT and
LRT availability
Utility functions are
equivalent in terms of
investment, system and
cost characteristics, as
well as socio-
demographics
14. The University of Sydney Page 14
PREFERENCE MODEL ESTIMATION
HMNL
1
, , ,
1,
, , ,
(1 )
...
(1
bus metro train
inv
bus metro train
HMNL
BRT fr BRTinv bus freq BRTinv metro freq BRTinv train
BRT x inv BRT
fr BRTsys bus freq BRTsys metro freq BRTsys tr
U freq freq freq
ASC x
freq freq freq
1 1
1
1, 1
, , ,
,
)
... ...
(1 )
sys
bus metro train
scst
ain
x sys BRT Z
fr BRTcst bus freq BRTcst metro freq BRTcst train
x cst BRT
x Z
freq freq freq
x
1
, , ,
_ , _
1,
(1
)
... ...
bus metro train
inv
HMNL
LRT freq inv bus freq inv metro freq inv train
usePT BRTLRT inv usePT BRTLRT
LRT x inv LRT
U freq freq freq
dummy
ASC x
Overt experience has a
difference influence on
the investment, system
and cost characteristics
Investment
characteristics
System characteristics
Cost
characteristics
Equivalent for LRT adding
the experience dummy
variable PT use with
availability for each
subset of attributes
15. The University of Sydney Page 15
COMPARISON OF THE MODELS – VUONG TEST
Vuong Test Prefer Metro Value Vote Live
HMNL vs
MNL
Vuong Statistic 1.825 0.508 1.698 1.283 2.257
Result
Favours HMNL
with 90%
confidence level
No statistically
significant
improvement
Favours HMNL
with 90%
confidence level
Favours HMNL
with 80%
confidence level
Favours HMNL
with 95%
confidence level
HMNL vs
HMNL0
Vuong Statistic 1.236 0.154 2.081 0.905 2.690
Result
Favours HMNL
with 75%
confidence level
No statistically
significant
improvement
Favours HMNL
with 95%
confidence level
No statistically
significant
improvement
Favours HMNL
with 99%
confidence level
Preferred Model HMNL
16. The University of Sydney Page 16
EMPIRICAL FINDINGS PREFERRED HMNL MODEL
– Gender was the only socio-demographic characteristic
which was significant in the personal preference
model (‘Prefer’) and the model associated with benefits
in the metropolitan area (‘Metro’).
– Female respondents were more inclined towards the
BRT alternative (often associated with safety in the
presence of a driver).
17. The University of Sydney Page 17
EMPIRICAL FINDINGS - DRIVERS
BRT LRT
Prefer Metro Value Vote Live Prefer Metro Value Vote Live
Investment Characteristics
Construction cost ($m) X X X X X X X X X X
Construction time (year) X X - X X X (A, UK) X X X -
Percent metro population serviced (%) X X X X X X X - X X
Percent right of way - X - X X X - X - X
Annual operating and maintenance cost ($m) X - X X X - X - - -
Operation period assured (year) X - X X X - X - - -
Risk of being closed after assured period (%) - X X X X X (P) X (P) - X (P) -
Environmental friendliness (% better/worse vs. car) X (P) - - X (P) X (P) X X X X X
Percent car switched to this mode (%) X (F) X X (F) X (F) X (F) - - X - -
High level of business attracted to station/stop (1/0) - - - - - X X X X X
System Characteristics
One-way service capacity ('1000 passengers) X X (A) - X (A) X (A) - - - - -
Off-peak headway (mins) X (A) X X - X (A) - - - - -
Travel time compared to car (% quicker/slower) X X - X X X X - X X
Travel cost compared to car (% cheaper/dearer) X X X X X X X X X X
Off-vehicle prepaid ticket required (1/0) X - - X (F) X (F) X - - - X
Integrated fare availability (1/0) X - X (A) - X (A) - X (A) - X -
Waiting time if transfer (mins) - - - - - X X X (US, P) X X
Staff presence on board (1/0) - - - - - X X X X X
Level boarding (vs. step boarding) - - - X - - - - - -
Total number of investment and system
drivers
13 10 9 14 15 12 13 9 11 10
18. The University of Sydney Page 18
EMPIRICAL FINDINGS - EXPERIENCE
Experience Prefer Metro Value Vote Live
Frequency bus conditioning investment characteristics X X - X X
Frequency rail conditioning investment characteristics X - - - -
Frequency metro conditioning investment characteristics X X X - X
Frequency bus conditioning system characteristics - X X - X
Frequency rail conditioning system characteristics X - - - -
Frequency metro conditioning system characteristics X X - X X
Frequency bus conditioning cost X - X - -
Dummy UsedPT_BRTLRT conditioning investment characteristics X - - - X
Dummy UsedPT_BRTLRT conditioning cost X - - - -
Total number of significant experience parameters 8 4 3 2 5
19. The University of Sydney Page 19
WILLINGNESS TO PAY ESTIMATES
Reduce construction time by 1 year Increase by 1% the population served Increase by 1% the right of way
0.00 2.00 4.00
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD $m)
Reduce the construction time in one year
BRT
LRT
0.00 1.00 2.00
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD $m)
Increase the population serviced by 1%
BRT
LRT
0.00 0.10 0.20
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD $m)
BRT
LRT
Increase the percentage right of way by 1%
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
20. The University of Sydney Page 20
WILLINGNESS TO PAY ESTIMATES
Increase by 1% the environmental
friendliness compared to car
Reduce travel time relative to car by
1%
Reduce travel cost relative to car by 1%
0.00 1.00 2.00
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD $m)
BRT
LRT
Increase the environmental friendliness in 1% compared to the car
0.00 0.20 0.40 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD $m)
BRT
LRT
Increase the travel time compared to the car by 1% quicker
0.00 0.20 0.40 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD $m)
BRT
LRT
Reduce the travel cost compared to the car by 1%
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
21. The University of Sydney Page 21
SIMULATED SCENARIOS
Level of support towards BRT in base scenarios for each
country:
Prefer 47.52% 47.44% 47.82% 49.57% 43.76%
Metro 46.97% 46.42% 49.03% 47.29% 47.21%
Value 49.24% 51.57% 49.28% 46.70% 48.90%
Vote 44.96% 46.17% 44.78% 46.12% 48.90%
Live 44.96% 46.17% 44.78% 46.12% 47.51%
Australia France Portugal U.K. U.S.A.
22. The University of Sydney Page 22
SIMULATED SCENARIOS – CONSTRUCTION COSTS
0%
2%
4%
6%
8%
%ofsupportchange
LRT construction cost is double
that of BRT
-15%
-10%
-5%
0%
%ofsupportchange
BRT construction cost is double
that of LRT
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
23. The University of Sydney Page 23
SIMULATED SCENARIOS – CONSTRUCTION TIMES0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
%ofsupportchange
LRT construction time is double
that of BRT
-2.5%
-2.0%
-1.5%
-1.0%
-0.5%
0.0%
%ofsupportchange
BRT construction time is double
that of LRT
24. The University of Sydney Page 24
SIMULATED SCENARIOS – CATCHMENT AREA0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
0.0%
0.5%
1.0%
1.5%
2.0%
%ofsupportchange
BRT serves 50% more people
-2.5%
-2.0%
-1.5%
-1.0%
-0.5%
0.0%
%ofsupportchange
LRT serves 50% more people
25. The University of Sydney Page 25
SIMULATED SCENARIOS – SUSTAINABILITY0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
-5.0%
-4.0%
-3.0%
-2.0%
-1.0%
0.0%
%ofsupportchange
LRT is 10% more
environmentally friendly
26. The University of Sydney Page 26
SIMULATED SCENARIOS – EXPERIENCE
-4.0%
-2.0%
0.0%
2.0%
4.0%
6.0%
%ofsupportchange
Bus frequency of use increases
100%
-4.0%
-3.0%
-2.0%
-1.0%
0.0%
1.0%
%ofsupportchange
Metro frequency of use
increases 100%
0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
27. The University of Sydney Page 27
SIMULATED SCENARIOS – AVAILABILITY0.00 0.10 0.20 0.30 0.40 0.50 0.60
Australia
France
Portugal
UK
US
Australia
France
Portugal
UK
US
WTP (AUD$)
Prefer Metro Value Vote Live
BRT
LRT
Reduce the travel cost compared to the car by 1%
-3.0%
-2.0%
-1.0%
0.0%
1.0%
2.0%
Australia U.S.
%ofsupportchange
All the cities have BRT and LRT
available
28. The University of Sydney Page 28
KEY FINDINGS
– Experience seems to have a significant influence
regarding overall performance of the model, and
the influence seems to vary across subsets of
attributes.
– However, the influence of experience on
preferences for BRT or LRT is small. This is an
important finding since it takes the pressure off the
much promoted position that until you experience
a mode you are unlikely to obtain sufficient
community support for it.
29. The University of Sydney Page 29
KEY FINDINGS
– Different measures of preference revelation
produce noticeable differences in the levels of WTP,
including different subsets of statistically
significant WTP estimates.
– There is clear evidence of preference heterogeneity
between countries between modal support drivers
and between modes, suggesting that replication as
a basis of transferability of evidence is potentially
problematic.
30. The University of Sydney Page 30
KEY FINDINGS
– The scenario simulations show some high
sensitivities in the levels of support for one mode
over the other.
– The greatest percent changes are associated with
construction cost, construction time, and
environmental friendliness.
31. The University of Sydney Page 31
KEY FINDINGS
– Historically, cost benefit analysis has used self-
interest or private consumer preference, and to avoid
any risk of double counting, this should remain the
appropriate metric to capture societal preferences in
CBA.
– The formal economic cost-benefit analysis can be
complemented by incorporating the preferences of
residents as expressed in a number of other ways.
– The results shows that the voice of residents has a
number of interpretations that might highly influence
the outcomes.
32. The University of Sydney Page 32
KEY FINDINGS
– This has the intent, amongst other reasons, of
drawing to the attention of politicians, their advisers
and the government bureaucracy, that the voice of
the residents has a number of interpretations that
have buy in appeal and will likely lead to a positive
electoral outcome, regardless of the CBA finding.
33. The University of Sydney Page 33
Do preferences for BRT and
LRT change as a voter,
citizen, tax payer, or self-
interested resident?
Camila Balbontin
David A. Hensher
Chihn Q. Ho
Corinne Mulley
June, 2018
BRT Centre of
Excellence webinar
Hinweis der Redaktion
19 cities in 5 countries
Does not fit in one column.
*We tested different combinations of availability and PT use, but this one was the only one that seemed significant.
*The table does not differentiate the influence on the BRT or LRT alternative, because if an experience parameter is included in either one of the alternatives, it would also be influencing the choice probability of both alternatives.
*Only US and Australia because they were the only two countries that had cities with both systems available.