The Impact of Bicycle-sharing on Conventional Commuting Travel
Structure
Jingyao Qu
1
,Wei Wang
2
, Qi Fan
1
and Mujie Lu
1
1
School of Transportation, Southeast University, Nanjing, Jiangsu, China
2
Jiangsu Key Laboratory of Urban ITS, Jiangsu Province Collaborative Innovation Center of Modern Urban Traffic
Technologies, School of Transportation, Southeast University, Nanjing, Jiangsu, China
Keywords: Bicycle-sharing, Multinomial Logit Model, Travel structure, RP survey.
Abstract: As a new travel mode, the emergence of bicycle-sharing (BS) can effectively solve “last mile” problem
which causes inconvenience in public transportation travel. More and more travelers choose using bicycle-
sharing instead of conventional travel modes. Previous researches about BS mostly were only based on the
large data, while the mechanism of how BS impacts the conventional travel structure can hardly know.
Multinomial Logit Model (MNL), a discrete selection model, can be used to compare the differences before
and after the emergence of BS. Based on the RP (revealed preference) survey results, the paper uses Stata
software to perform the logit analysis about the travel mode choice of travelers. The main factors impact
travel model choices are selected and parameters are also calculated. Finally, the utility function of each
travel modes are calculated. The results are compared, providing a reference for future traffic planning and
the adjustment of traffic management policy.
1 INTRODUCTION
Advantages like flexible parking points, little
limitation for users, economic security and payment
convenience make the sharing bikes become the new
trend of people’s travel choice.
The emergence of bicycle-sharing (BS) leads to a
solution of “last mile” problem with its flexible
parking property. The emergence of BS still has a
huge impact on conventional travel structure. From
the "Bicycle-sharing and Urban Development White
Paper in 2017", it can be seen exactly that the
emergence of sharing bicycles directly leads to a
strong increase of bicycle travel in city. It also leads
to a significant reduction on car travel, especially
unlicensed cabs.
As a new mode of traffic travel, there is
relatively little research about bicycle-sharing. NPV,
IRR model are used to analyze the profitability of
shared bicycle companies with ofo and moblike as
examples (Li, 2017). The characteristics and
functions of sharing bikes are analysed (Wang,
2017).
Plenty of researches were done about public
bicycle, a traffic mode which is relatively similar to
the BS pattern. A survey is always carried out before
travel satisfaction analysis (Liu, 2016) and travel
mode selection analysis (Shaheen SA, 2013; Zhu,
2012). On this base, the factors that influence
likelihood of using public bicycles and frequency are
analyzed (Bachand-Marleau, J, 2012; Cao, 2015;
Shaheen, SA, 2011). Methods like discrete choice
model (Shen, 2015; Luo, 2013), Fuzzy
comprehensive Evaluation, empirical analysis of
tour-based bicycle use, analysis of IC card data
(Cao, 2016) and difference-in-differences regression
model (Kayleigh B. Campbell, 2017) are used for
traffic needs analysis and to obtain the general
proportion of public bicycles in traffic structure.
To conclude, though there are many studies
about BS, these studies are limited to a summary of
sharing bikes’ large data. However, as a new mode
of traffic travel, the study must be proceeded from
the analysis of traffic demand. In this way, the travel
mechanism of BS can be analyzed, which cannot be
obtained from large data. As a similar mode of
traffic travel with BS, the research methods of public
bicycles can be used to analyze the demand
mechanism of BS.
258
Qu, J., Wang, W., Fan, Q. and Lu, M.
The Impact of Bicycle-sharing on Conventional Commuting Travel Structure.
In 3rd International Conference on Electromechanical Control Technology and Transportation (ICECTT 2018), pages 258-264
ISBN: 978-989-758-312-4
Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

Thus, the research goals can be determined:
(1)The paper takes the travel structure of Nanjing
residents as research objective, then uses RP survey
to investigate the modes of residents’ travel, the
results of investigation can be used to identify the
main factors that affect the choice of traffic travel.
(2)With the data of the investigation, the paper use
Stata software to establish the MNL model of
residents’ travel mode choices before and after the
emergence of BS. (3)The paper compares the
differences of travel modes choices before and after
the emergence of BS, and gives suggestions to
government and enterprises for better traffic
planning and management.
The rest of the paper is organized as follows:
Section 2 describes survey design and data
collection process. Section 3 discusses sample
characteristics before and after the emergence of BS.
Section 4 shows modelling process and results
discussion. Conclusions and future directions are
provided in section 5.
2 SURVEY DESIGN AND DATA
COLLECTION
2.1 Research Objective
As a new way of traffic travel, the emergence of BS
has a huge impact on conventional travel structure,
which can be studied to make cities’ traffic planning
and traffic managements. Thus, the paper takes the
travel structure before and after the emergence of
sharing bikes as study objective to explore the
differences between two conditions.
As a similar way of traffic travel with BS, the
methods used to investigate public bicycle travel can
be used to investigate the travel mode of BS. The
sharing bicycles are upgraded version of public
bicycles. Similarly, the sharing bicycles are more
economically safe and flexible than private bicycles,
as a result, it must has a huge impact on
conventional travel structure, which is a critical
factor to traffic planning and management.
2.2 Survey Design
The questionnaire adopts the method of RP survey
(Revealed Preference survey), mainly considering
the impact of the personal attributes, transfer
characteristics and perceptions of travel satisfaction.
Consisting following parts:
(1) Personal attributes: including sex, age,
profession, income level, private transport condition,
which may affect the mode choice of traffic travel
and different people have different travel factors
about personal attributes.
(2) Travel characteristics: including travel form
and purpose, main travel time period, travel
distance, the choice of travel mode, travel expense
and travel time consuming. The choice of travel
mode can be regarded as the dependent variable of
the travel structure study, and other travel
characteristics are factors that affect travel mode
choosing.
(3) Perceptions of travel satisfaction: including
travel considerations, attitude towards public
transportation travel, the main reasons that affect the
public transportation travel, attitude towards travel.
The travel considerations including transfer
convenience, travel safety, travel punctuality, green
travel, travel comfort and travel expense. The main
reasons that affect the public transportation travel
including waiting time, traffic transfer conditions,
travel comfort, travel speed and expense. The travel
attitude values are varied from very dissatisfied to
very satisfactory.
(4) The travel characteristics section of the
questionnaire is designed in two parts, travel
investigation before the emergence of BS and after
the emergence of BS.
2.3 Filed Survey and Data Collection
The survey was carried out in different areas in
Nanjing, and was carried out concretely near the
public transport sites and transport hub in August,
2017, and received 487 valid case, among them, 415
questionnaires are valid questionnaires.
3 DESCRIPTIVE ANALYSIS
The survey covered the traffic travel before and after
the emergence of BS.
Sample characteristics are
explored as follows.
3.1 Passenger Personal Attributes
As shown in Table 1, 53.5% of the respondents are
men and 46.5% of the respondents are women, both
of which are close to the theoretical value 50%.
Over 90% of the respondents’ age are between 18
and 50, this is consistent with the age distribution of
commuter travelers in the actual situation. Over 50%
of the respondents are workers and nearly 40% are
The Impact of Bicycle-sharing on Conventional Commuting Travel Structure
259

students or teachers, all of which are the main forces
of the commuting travelers. The distribution of the
income is consistent with the actual situation.
Table 1: Passenger personal attributes distribution
proportion.
Personal attributes Proportion
Sex
Male
Female
53.5%
46.5%
Age
9-18
18-30
30-40
40-50
>50
3.9%
42.9%
30.6%
20.2%
2.4%
Profession
Student
Teacher
Enterprise or government
staff
Individual business
household
Service worker
Others
21.9%
16.6%
33.3%
9.4%
8.9%
9.9%
Private
transport
condition
Private car
Private bicycle
Private electric bicycle
No private transpor
t
31.3%
20.5%
16.4%
49.6%
Income
<1500
1500-3000
3000-5000
5000-8000
8000-12000
>12000
21.7%
18.1%
32.5%
14.7%
7.0%
6.0%
Note: Private transport condition has multiple
options, so the total probability not equal to 1.
3.2 Travel Characteristics
For commuting travel, the travel characteristics are
shown in Table 2. Nearly 2/3 of the respondents’
purposes are working and most of their travel time
period are distributed in the morning and evening
peak hours. Compared with the situation that before
the emergence of BS, the travel mode choices of car
have significantly reduced and the choices of public
transport have significantly improved. The largest
increase in travel mode choice is bicycle travel. All
of the results shows that the emergence of BS are
beneficial for development of public transportation
and protecting environment.
Table 2: Travel characteristics of commuting travel.
Travel characteristics
Before
(%)
After
(%)
Travel
purpose
Working
Go to school
Business
othe
r
68.4
18.6
8.4
4.6
Travel
time
(period)
0:00-7:00
7:00-9:00
9:00-14:00
12.8
68.0
14.9
14:00-17:00
17:00-19:00
19:00-24:00
16.6
45.3
25.3
Travel
distance
<500m
500-1000m
1000-2000m
2000-4000m
4000-7000m
7000-10000m
>10000
m
5.5
17.3
25.1
20.5
15.2
6.0
10.4
Travel
mode
Private car
Taxi
Bus & Walk
Bus & Public bicycle
Bus & Private bicycle
Bus & Sharing bicycle
Metro & Walk
Metro & Public bicycle
Metro & Private bicycle
Metro & Sharing bicycle
Walk
Public bicycle
Private bicycle
Sharing bicycle
27.7
16.6
37.8
11.6
8.9
0
25.1
4.8
2.7
0
17.3
6.5
13.5
0
25.5
11.3
25.5
12.8
7.0
22.9
12.3
4.8
2.0
17.8
16.4
4.8
13.3
20.7
Expense(
Yuan)
0
0-100
100-200
200-500
500-1000
>1000
8.7
37.1
24.6
18.3
9.9
1.4
7.2
41.9
23.9
18.3
7.0
1.7
Time
consumin
g(min)
0-10
10-20
20-30
30-40
40-60
>60
8.2
30.4
28.2
18.8
9.2
5.2
11.6
31.3
29.6
15.4
8.7
3.4
Note: Travel time (period) and travel mode have
multiple options, so the total probability not equal to
1.
3.3 Travel Characteristics
Before the emergence of BS, 11.1% of commuting
travel respondents are very dissatisfied with public
transport travel, 16.4% are dissatisfied, 41.9%
feeling okay with it, 20.7% are satisfied and 9.9%
are very satisfied. The most influential factor in
public transport travel is long waiting time (70.8%).
After the emergence of BS, only 14.5% of
commuting travelers are dissatisfied or very
dissatisfied with public transport travel, a significant
reduction compared with the proportion before
(27.5%). The proportion of inconvenient transfer has
a significant reduction, with 37.3% compared with
54.7%. The data reveals that BS is beneficial for
development of public transport travel.
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
260

4 MODELLING PROCESS AND
RESULT DISCUSSION
In this study, the dependent variable y is multiple
dependent variable. Before the emergence of BS, the
option set is set to a number set changing from 1 to
11 with changing of mode choices. After the
emergence of BS, the option set is set to a number
set changing from 1 to 14. Thus, the multinomial
logit model is appropriate. The independent
variables for before A and after B models are defined
as x
a1
, x
a2
… x
an
and x
b1
, x
b2
… x
bn
which are factors
of personal attributes, travel characteristics and
perceptions of travel satisfaction. Taking commuting
travel before the emergence of bicycle-sharing for
example, the multinomial logit model can be
expressed as:
1
in
jn jn in
nn
V
in
VVV
jA jA
e
P
ee




iA
(1)
where P
in
is the probability of traveler n selects
travel mode i; V
in
is the fixed item in the utility
function of traveler n selects travel mode i, taking
the linear function of the parameter vector
and the
eigenvector X
in
; A is the travel mode choices set
before the emergence of BS.
The utility of the alternatives to the traveler can
be expressed in the form of the following functions
：
in ij aij i
j
Vxc  

(2)
where α
ij
are explanatory variable coefficients, x
aij
is
independent variables that traveler choose mode i, c
i
is the inherent dummy variable of traffic i.
4.1 Independent Variables Selection
It is necessary to eliminate factors with less impact
on mode choice before modelling. Due to the
multiple dependent variable, maximum likelihood
ratio test are used to filter variables, the final
independent variables are shown in Table 3. For lack
of space, we only show the variables of commuting
travel:
Table 3: Independent variables and pretreatment before
modelling.
Types Variables Description
Personal
attributes
Sex: Male
Sex: Female
Age
Income
Private transport
condition(car)
Private transport
condition(electric bicycle)
Private transport
condition(bicycle)
0
1
1,2,3,4,5
1,2,3,4,5,6
1,0
1,0
1,0
Travel
characterist
ics
Travel purpose
Travel distance
Expense
Time consuming
1,2,3,4
1,2,3,4,5,6,7
1,2,3,4,5,6
1,2,3,4,5,6
Perceptions
of travel
satisfaction
Considerations(transfer
convenience)
Considerations(safety)
Considerations(punctuality)
Considerations(environmenta
l protection)
Considerations(comfort)
Travel satisfaction level
1,0
1,0
1,0
1,0
1,0
1,2,3,4,5
4.2 Model Calibration Results
The model is constructed in Stata. As Table 4 shows,
most parameter estimates are significantly at 90%
confidence levels with expected sign, which means
most of passengers’ personal attributes, travel
characteristics and perception of travel satisfaction
are main factors for travel mode choice. The
estimation results are shown in Table 4, for lacking
of space, only some results before the emergence of
BS are shown, other results are also shown in
statistical analysis part:
The Impact of Bicycle-sharing on Conventional Commuting Travel Structure
261

Table 4: Estimation results of commuting travel before.
Travel
m
ode Age Private-ca
r
Private car
taxi
Bus & walk
Bus& public
bicycle
Bus & private
bicycle
Metro& walk
Metro& public
bicycle
Metro& private
bicycle
Walk
Public bicycle
~
~
~
~
~
-0.509(0.084
-1.026(0.054)
~
~
~
4.583(0)
1.334(0.030)
~
~
~
~
~
~
~
~
Private-
b
icycle purpose distance
-2.664(0)
-2.767(0)
-2.456(0)
-2.434(0)
-1.929(0.002)
-4.032(0)
-3.236(0.006)
~
-2.238(0)
-1.145(0.077)
0.537(0.086)
~
0.495(0.073)
~
~
~
~
0.701(0.099)
~
~
0.319(0.080)
0.303(0.084)
0.288(0.081)
~
~
0.486(0.006)
~
~
-0.376(0.048)
~
environmental satisfaction constan
t
~
~
~
1.208(0.056)
~
~
~
1.597(0.096)
1.410(0.021)
1.760(0.012)
~
0.543(0.017)
0.516(0.015)
0.713(0.004)
0.638(0.017)
~
0.977(0.003)
0.949(0.018)
0.537(0.023)
0.666(0.020)
-3.200
-2.533
-1.829
-3.535
-3.954
~
-6.414
-5.764
~
-3.941
LL(0) -1433.0822
LL( )
-930.14745
adj.
2
0.253
Note: The numbers in the box are coefficients,
and the numbers in parentheses are significances.
The value of Mc Fadden’s adj R squre is 0.253,
the value is in the range of 0.2 to 0.4, indicating that
the model fitted well.
4.3 Statistical Analysis of Travel Mode
Choice
According to the established utility functions of
different travel modes, the probability functions of
different travel modes can also be obtained. The
utility functions of each modes are shown as Table
5. For lacking of space, only some results are shown.
Table 5: The utility functions of commuting travel modes
before and after.
Modes
Before/
afte
r
Utility function
Car
B
cov
4.583 1.559 2.664
0.537 0.319 2.542
1.055 3.200
car
p
car peb pb
prp dis
cmf
V
x
xx
x
xx
x




A
3.774 0.492 1.751
0.519 0.845
car
p
car dis enp
sati pun
V
x
xx
xx



Taxi
B
cov
1.334 1.586
2.767 0.303 2.585
0.543 2.533
taxi pcar peb
pb dis
sati
Vx x
x
xx
x

 

A
cov
1.084 0.436
1.237 0.512 1.238
0.879 5.534
taxi sex age
pb prp
cmf
Vxx
x
xx
x

 

B& priB
B
&
cov
1.307 1.929
3.123 0.638 3.954
B
priB peb pb
sati
Vxx
xx
 
 
A
&
0.415 1.370
BpriB dis
Vx
B& B-S
B
—
A
&
cov
0.255 0.409 1.052
BBS
inc dis
V
x
xx


 
M& priB
B
&
cov
3.176 0.701 1.766
1.597 0.949 5.764
MpriB
peb prp
enp sati
V
x
xx
xx


 
A
&cov
1.653
MpriB
Vx
M& B-S
B
—
A
&
cov
1.641 0.556
0.788 1.385 4.672
BBS
pcar prp
dis
V
xx
xx



 
Private
bicycle
B
—
A
0.591 1.056
2.129 2.085 1.539
0.585
priB age pcar
p
eb pb enp
sati
Vxx
x
xx
x



Bicycle-
sharing
B
—
A —
It can be seen from the Table 5, before the
emergence, the coefficient of private car ownership
condition for private car travel is 4.583, compared
with 3.774 after the emergence, which reveals the
fact that after the emergence of BS, fewer and fewer
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
262

travelers choosing car for travel, even they have
private cars. It means that BS is beneficial for
reducing car travel. Compared with the before
condition, the coefficients of satisfaction become
negative from positive, it means the travelers’
options are no longer violently fluctuating, which is
a sign that the emergence of BS is a trend to
narrowing the service gap between the different
travel modes. Taking Metro& Walk modes for
example, the coefficient of private bicycle
ownership condition are changed from -4.032 to -
2.372, it means that the effect of this ownership
condition for choosing of public transport are
reducing, this reflects the fact that the emergence of
shared bicycles is conducive to the development of
public transport.
5 CONCLUSIONS
The paper takes the travel structure of Nanjing
residents as research objective. A carefully designed
survey was conducted to capture the travelers’ travel
mode choices before and after the emergence of BS.
Then the survey results were analyzed and variables
were selected. The paper uses Stata to establish a
MNL model for travel choice prediction. After
comparing the different travel choices model before
and after the emergence of BS, following
conclusions about the impact of BS to conventional
travel structure are obtained:
Firstly, BS has a huge impact on car travel. It not
only directly takes away some original car travellers,
but also improves the transfer condition, It improves
the roadway utilization efficiency and reduces
pollution emission.
Secondly, BS travel also has a huge impact on
public transport travel. For bus, BS is playing a
competitor role, this is because the speed of bus is
relatively slow and buses are often stuck in the
traffic congestion during the peak hours. For metro
travelers, BS is playing a role of assistant. Most
metro users are middle or long distance travelers, at
this point, sharing bicycle travel almost has no
impact on metro travel. Sharing bicycle travel is also
a good solution for the trip from the starting point to
subway station. In summary, the impact of shared
bicycles on public transport is multifaceted, but
overall is playing a positive role to public transport
travel.
Several future directions can be proposed based
on this study. Firstly, the impact of BS to public
transport travel is based on the travel mode, the
result of competition between BS and bus seem to be
inconclusive. An issue about the relationship
between BS and bus based on distance can be a new
extension to the paper. Secondly, the results of the
model told us that the emergence of BS is a sign that
the service gap between the different travel modes
are narrowing, the mechanism of this phenomenon
can be another extension to the paper.
ACKNOWLEDGEMENT
This research is supported by National Natural
Science Foundation of China (51338003).
REFERENCES
Bachand-Marleau, J., Lee, BHY., El-Geneidy, AM., 2012.
“Better Understanding of Factors Influencing
Likelihood of Using Shared Bicycle Systems and
Frequency of Use.” Journal of the Transportation
Research Board, No. 2314, Washington, D.C., 2012,
66–71.
Cao, X. N. (2016). “Research on Operating Characteristics
and Dynamic Scheduling Optimization of Public
Bicycle Based on IC Card Data”. Master's Thesis,
Southeast university, Nanjing, China.
Cao, X. N., Wang, W., Ji, Y. J., Zhao, L. N. (2015).
“Influence Factors of Transfer Behavior Between
Urban Rail Transit and Public Bicycle.” Journal of
Transportation Engineering and Information, 4, 96-
101.
Kayleigh, B., Campbell, Candace B. (2017). “Sharing
riders: How bikesharing impacts bus ridership in New
York City.” Transportation Research Part A, 100,
264-282.
Li, M. L. (2017). “Research and Analysis of Shared
Bicycle Market.” Finance and Economics (Academic
Edition), 5, 121-123.
Liu, B. (2016). “Research on Evaluation of Public Bicycle
Satisfaction Based on Fuzzy Comprehensive
Evaluation Method”. Master's Thesis, Chang’an
university, Shannxi, China.
Luo, B. (2013). “Study on Urban Public Bicycle Choice
Behavior”. Master's Thesis, Chang’an university,
Shannxi, China.
Shaheen, S. A., Martin, E. W., Adam, P. C. (2013) “Public
Bikesharing and Modal Shift Behavior: A
Comparative Study of Early Bikesharing Systems in
North America.” International Journal of
Transportation, 1, 35-54.
Shaheen, S. A., Zhang, H., Martin, E., Guzman, S. (2011).
“Hangzhou Public Bicycle: Understanding Early
Adoption and Behavioral Response to Bikeshare in
Hangzhou, China.” 2011 Transportation Research
Board Annual Meeting.
The Impact of Bicycle-sharing on Conventional Commuting Travel Structure
263

Wang, G. R. (2017). “Traffic value and development path
of the bicycle-sharing.” City traffic, 4, 72-75.
Shen, X. (2015). “Study on the Locating and Planning of
Public Bike Rental Station Based on the Transfer with
Urban Rail Transit”. Master's Thesis, Chang’an
university, Shannxi, China.
Zhu, W., Pang, Y. Q., Wang, D., Yu, X. W. (2012).
“Travel Behavior Change after the Introduction of
Public Bicycle Systems: A Case Study of Minhang
District, Shanghai.” Journal of Urban Planning, 5, 76-
81.
ICECTT 2018 - 3rd International Conference on Electromechanical Control Technology and Transportation
264