Presentation by Jan-Dirk Schmöcker of of Kyoto University. Delivered at the Institute for Transport Studies (ITS), 27 November 2014.
http://trans.kuciv.kyoto-u.ac.jp/its/Schmoecker.html
Model Call Girl in Rajiv Chowk Delhi reach out to us at 🔝9953056974🔝
Estimation of positive demand feedback processes
1. Estimation of Positive Demand
Feedback Processes
Jan-Dirk Schmöcker
schmoecker@trans.kuciv.kyoto-u.ac.jp
2. Demand Dynamics
Often initial demand for new transport schemes is
lower than predicted.
However, some systems experience a sudden
“demand boom” after some time.
Possible to predict these?
Service quality improvements often do not seem to
2
4. Demand Factors
Service quality
Travel needs
Attitudes, Social Norms, Perceptions:
Dynamics due to “Adaptation”
4
5. Social Norms
5
Descriptive: “The majority is probably right”
Injunctive: “Do what is expected of one”
Often a combination of both influence behaviour
Positive feedback, hence possibly lead to: Mass effect,
fashion, trend, bandwagon effect, snowball effect,…
Numerous examples, including travel behaviour
6. To distinguish different types of causes for user adaptation
(Schmöcker, Watling and Hatori, 2013):
■ “Real effects” (congestion, traffic safety)
■ Consequential effects (economies of scale)
■ Perceived effects
■ Information effects, including social norms
Are these effects distinguishable and if yes, quantifiable?
Modeling as well as data question: Focus in this
presentation on empirical approaches
6
Resulting research questions
8. 8
Car ownership desire
modelling
Survey among students how strongly they desire to
purchase a car in the future
Questionnaire in 7 countries:
Japan, NL, US, Beirut, Shanghai, Taiwan, Indonesia
9. 9
Questionnaire design
Dependent variable: intentions to buy a car in the future
(next 10 years) measured on a 7-point Likert scale (very
unlikely – very likely)
Explanatory latent variables:
■ Expectation of others to buy a car
■ Symbolic affective attitudes towards cars
■ Perceptions that the car brings “Independence”
Explanatory observed variables:
■ Regular car use, Personal Income
Further socio-demographic variables and other attitudinal
factors also asked but not found significant in the
subsequent analysis
10. 10
Latent variables and indicators
Variables Indicators Measurement
zex Expectations
Of Others
I1 Parents 1 = They strongly expect me not to buy a car;
2 = They expect me not to buy a car;
3 = They have a little bit expectation of me not
to buy a car;
4 = They have no expectation;
5 = They have a little bit expectation of me to
buy a car;
6 = They expect me to buy a car;
7 = They strongly expect me to buy a car
I2 Partner
I3 Family members and relatives
I4 Close friends
I5 Peers at university
I6 People in neighborhood
I7 People in province/state
zsy Symbolic
affective
I8 Cars are cool
1 = strongly disagree;
2 = disagree;
3 = somewhat disagree;
4 = neutral;
5 = somewhat agree;
6 = agree;
7 = strongly disagree
I9 Cars allow to express oneself
I10 Cars are trendy
I11 Cars bring prestige
I12 Cars allow to distinguish oneself from
others
I13 Cars are fun to have
zin Indepen-
dence
I14 Cars are convenient
I15 Cars allow one to travel anytime
I16 Cars allow one to be independent
I Cars allow one to travel anywhere
11. 11
Model Framework
(Ordered Hybrid Latent Choice Model)
Utility
y: intention to buy a
car (1-7)
I8: Cars are cool
I9: Cars allow to
express oneself
Symbolic
Affective
I10: Cars are trendy
I11: Cars bring
prestige
I12: Cars allow to
distinguish oneself
from others
I13: Cars are fun to
have
Regular Car Use Personal Income
Missing Income
Dummy
Independence
I14: Cars are
convenient
I15: Cars allow one
to travel anytime
I16: Cars allow one
to be independent
I17: Cars allow one
to travel anywhere
Expectation of
Others
I1: Parents
I2: Partner
I3: Family members
& Relatives
I4: Close friends
I5: Peers at
University
I6: People in
neighborhood
I7: People in
province/state
12. Main Results
All variables mentioned are significant with expected signs
Personal Income and Regular Car Use positively influence
car purchase Intention
Expectation of Others has larger influence on Car Purchase
intention than both Symbolic Affective and Independence
Parents most important, close friends and peers have similar
influences, whereas the wider population (people in
neighborhood and state) have relatively lower influences.
Expectations of peers is perceived more uniformly,
expectation of parents has more variability.
12
13. Extended Framework (on-going work)
Interaction of “Expectation of Others” with
“Importance of Others”
■ …strong expectations are not important if I don’t care about
the person or group!
Better approximation of “Social Norms”
Different model specifications possible
■ Group specific interactions of expectations and influence
■ Then as latent or exogeneous variables
■ Interact total “influencability” and total perceived expectations
of a person
13
20. Station Specific Models
Further models fitted for ridership from/to specific
stations
Accessibility to stations important, but possibly there
exists a “threshold”, i.e. it appears the “first
connection” is much more important than subsequent,
additional ones.
Adaptation effects vary depending on location of
station and with it predominant mode of travel of
travellers from the station.
20
21. Taipei and Hsinchu case
21
0
500000
1000000
1500000
2000000
2500000
2007 2008 2009 2010 2011 2012
Taipei ridership Taipei prediction
Hsinchu ridership Hsinchu prediction
Model: Loglinear with MA(1) Taipei Hsinchu
Parameters lag Coeff. t Coeff. t
Total Population
0 529.76 2.14 -82.07 -1.23
1 551.10 2.19 -70.69 -1.07
Unemployment Ratio 3 -0.48 -4.95 -0.44 -3.15
GDP/Fuel Price 1 -0.21 -1.80 -0.39 -3.13
Car Ownership 0 0.31 1.24 0.09 0.37
Chinese New Year 0 0.01 0.39 -0.05 -2.09
Summer Vacation 0 0.06 2.73 0.07 2.84
Adaptation Effects 0 0.61 12.45 0.95 13.66
Constant 346.89 4.54 168.29 4.91
Observation 70 72
R-square 0.95 0.98
Adjusted R-square 0.94 0.97
Moving Average Coeff. -0.21 -1.51 -0.52 -4.44
Ljung-Box Q test 0.11 0.00
22. Partial Conclusions
Further models fitted for ridership from/to specific
stations
■ Accessibility to stations important – to some degree.
■ Adaptation effects vary depending on location of station and with
it predominant travel purpose from/to station.
Reaching an equilibrium takes time for new transport
schemes
Initial (and still) low demand has various causes, among
which one might be adaptation effects.
These could include various effects which aggregate
data wont tell.
22
24. Current analysis: Traveller Surveys
24
1. Hurt’s Scale of
Innovativeness (1977)
Measuring one’s “willingness-to-change” (Rogers, 1983)
2. Annual frequency and trips
purpose
1. General questions of HSR usage for respondents
(hopefully to helping them recognize their travel
pattern)
2. Recalling questions
3. Dynamic travel
pattern identification
10 visualized hypothetical long-term travel
pattern
4. Understanding motivations
for change/no change
5-level Likert scale questions
5. Most frequently used OD
and alternative mode
As a proxy to estimate travel time / distance
6. Personal info Gender, Marital status, Age, Average monthly budget,...
25. Usage Patterns and Distribution
25
ABC ABD A
ABC
AD
ABC
ABCD
ABC
AB
ABDB
Pattern Taiwan Shanghai
No. of
respondent
%
No. of
respondent
%
1 34 10.49% 51 11.83%
2 28 8.64% 24 5.57%
3 53 16.36% 53 12.30%
4 13 4.01% 59 13.69%
5 58 17.90% 37 8.58%
6 15 4.63% 21 4.87%
7 21 6.48% 25 5.80%
8 66 20.37% 95 22.04%
9 15 4.63% 25 5.80%
10 15 4.63% 34 7.89%
None of above 6 1.85% 7 1.62%
Total 324 100% 431 100%
26. Very initial survey observations
In both Shanghai and Taiwan we find that about 20%
of travellers immediately adapted.
Frequency of travellers increases with time in both
Shanghai and Taiwan
Most respondents start using HSR in the initial 3-4
years of service.
Percentage of business travellers especially in
Taiwan increased -> business adaptation?
Motivations to start using the system and frequency
patterns (change in frequency!) to some degree
related
26
27. Conclusions and Questions (1)
On individual level we can estimate the importance of
social norms and positive feedback process
On aggregate level we can also see positive
feedback process
■ Though not clear which kind of mass effect it is
■ Possibly the survey analysis helps?
27
28. Conclusions and Questions (2)
Still open question how to transfer the discrete choice
model results from Study 1 into aggregate demand
forecasting as in Study 2
How well can we (ever) calibrate agent-based models
and analytical approaches?
Maybe for the foreseeable future we can only do
“reverse modeling”, i.e. show the range of different
explanations that fits a current situation?
■ …and then extrapolate to the future with very stochastic
approaches
28
29. Papers related to this presentation
Abou-Zeid, M., Schmöcker, J.-D, Belgiawan, P.F. and Fujii, S. (2013). Mass Effects and
Mobility Decisions. Transportation Letters, 5(3), 115-130.
Belgiawan, P. F., Schmöcker, J.-D. and Fujii, S. (In press). Understanding Car Ownership
Motivations among Indonesian Students. Accepted for publication in International Journal
of Sustainable Transportation.
Belgiawan, P.F., Schmöcker, J.-D., Abou-Zeid, M., Walker and Fujii, S. (2015). The Role of
Expectation of Others on Students’ Likelihood to Buy a Car. Accepted for Publication at
the 94st Annual Meeting of the Transportation Research Board. Washington D.C., U.S.
Belgiawan, P.F., Schmöcker, J.-D., Abou-Zeid, M., Walker, J., Lee, T.-L., Ettema, D. and
Fujii, S. (2014). Car Ownership Motivations among Undergraduate Students in China,
Indonesia, Japan, Lebanon, Netherlands, Taiwan, and U.S.A. Transportation, 41, 1227-
1241.
Schmöcker, J.-D., Hatori, T. and Watling, D. (2014). Dynamic Process Model of Mass
Effects. Transportation, 41(2), 279-304.
Yeun-Touh, L., Schmöcker, J.-D. and Fujii, S. (In Press). Demand Adaptation towards
New Transport Modes: Case of High Speed Rail in Taiwan. Accepted for Publication in
Transportmetrica B.
29