ArticlePDF Available

Impact of e-scooter sharing on bike sharing in Chicago

Authors:

Abstract and Figures

As a new type of shared micromobility, e-scooter sharing first appeared in the United States and became popular worldwide. Considering e-scooter sharing and bike sharing have similar service attributes, the ridership of bike sharing may be affected by the introduction of e-scooter sharing. To date, studies exploring this impact are limited. In this study, we seek to analyze the impact of e-scooter sharing on the usage of bike sharing from trip data of e-scooter sharing and bike sharing in Chicago for a total of 30 weeks. We rely on a difference-indifferences modeling approach based on the propensity score matching method. We found that the average duration of e-scooter trips is shorter than that of bike trips. The introduction of e-scooter sharing reduced the overall bike sharing usage by 23.4 trips per week per station (10.2%). bike sharing usage of non-members and members decreased by 18.0 (34.1%) and 5.4 (4.0%) trips, and that of male and female members decreased by 3.3 (3.1%) and 2.0 (7.3%) trips, respectively. Furthermore, the volume of short-, medium-, and long-duration trips of bike sharing decreased by 10.9 (7.5%), 5.4 (9.6%), and 3.4 trips (20.5%), respectively. Finally, bike sharing use during non-peak hours decreased but was not affected during peak hours.
Content may be subject to copyright.
1
Impact of e-scooter sharing on bike sharing in Chicago
Hongtai Yang a, Jinghai Huo b,*, Yongxing Bao b, Xuan Li b,
Linchuan Yang c, Christopher R. Cherry d
a School of Transportation and Logistics, National Engineering Laboratory of Integrated Transportation Big
Data Application Technology, National United Engineering Laboratory of Integrated and Intelligent
Transportation, Institute of System Science and Engineering, Southwest Jiaotong University, Chengdu, China
b School of Transportation and Logistics, National Engineering Laboratory of Integrated Transportation Big
Data Application Technology, National United Engineering Laboratory of Integrated and Intelligent
Transportation, Southwest Jiaotong University, Chengdu, China
c School of Architecture and Design, Department of Urban and Rural Planning, Southwest Jiaotong University,
Chengdu, China
d Civil and Environmental Engineering, University of Tennessee-Knoxville, Knoxville, TN 37996-2313, USA
To cite: Yang, H., Huo, J., Bao, Y., Li, X., Yang, L., & Cherry, C. R. (2021). Impact of e-scooter
sharing on bike sharing in Chicago. Transportation Research Part A: Policy and Practice, 154,
23-36.
Abstract: As a new type of shared micromobility, e-scooter sharing first appeared in the United
States and became popular worldwide. Considering e-scooter sharing and bike sharing have similar
service attributes, the ridership of bike sharing may be affected by the introduction of e-scooter
sharing. To date, studies exploring this impact are limited. In this study, we seek to analyze the
impact of e-scooter sharing on the usage of bike sharing from trip data of e-scooter sharing and
bike sharing in Chicago for a total of 30 weeks. We rely on a difference-in-differences modeling
approach based on the propensity score matching method. We found that the average duration of
e-scooter trips is shorter than that of bike trips. The introduction of e-scooter sharing reduced the
overall bike sharing usage by 23.4 trips per week per station (10.2%). bike sharing usage of non-
members and members decreased by 18.0 (34.1%) and 5.4 (4.0%) trips, and that of male and
female members decreased by 3.3 (3.1%) and 2.0 (7.3%) trips, respectively. Furthermore, the
volume of short-, medium-, and long-duration trips of bike sharing decreased by 10.9 (7.5%), 5.4
(9.6%), and 3.4 trips (20.5%), respectively. Finally, bike sharing use during non-peak hours
decreased but was not affected during peak hours.
Keywords: Micromobility, Shared E-scooter, Bike sharing, Difference-in-differences, Propensity
score matching
2
1. Introduction
E-scooter sharing has developed rapidly worldwide since 2017. In 2019, shared e-scooter
trips doubled from 2018 and now make up 2/3 of all shared micromobility trips in the US (NACTO,
2020). Nearly one-fourth of Parisians used shared e-scooters in 2019 (Lime, 2019a). In Los
Angeles, the e-scooter sharing operator Lime provided 3 million trips from 2017 to the beginning
of 2019 (Lime, 2019b). In Austin, the cumulative number of shared e-scooter trips reached 5
million in July 2019, greatly exceeding that of bike sharing trips (Austin, 2019). Being dockless
means that an e-scooter can be parked anywhere within the service area instead of being returned
to the designated stations. Moreover, as a powered vehicle, e-scooters can be used to travel for
longer distances with less physical effort. Both features reduce some of the barriers associated with
bike sharing systems. The growth of e-scooter sharing has been fueled by private investment and
is faster than bike sharing. Notably, bike sharing has a longer history, and approximately 1,000
cities worldwide have bike sharing systems. The appearance of e-scooter sharing has brought a
new micromobility option to urban travel but the role of e-scooter sharing in urban transportation
systems is still unclear (Tuncer et al., 2020). In particular, e-scooter sharing and bike sharing are
both shared micromobility modes and are usually used for short-distance travel. Thus, they are
likely to compete for riders. It is important for government agencies to understand how the advent
of ESS influences the ridership of bike sharing so they find ways to coordinate the two systems to
better serve the public and expend public resources. Some examples include relocating bike
sharing stations, adjusting the capacities of different stations, pricing, and determining the e-
scooter sharing operation areas. This topic has not been explored in depth by previous studies. In
this study, we aim to quantify the impact of e-scooter sharing on the ridership of bike sharing.
In our study, the data of Chicago is adopted and analyzed because this city has both e-scooter
sharing and bike sharing systems and the trip data of both systems are open to the public. The e-
scooter sharing pilot program in Chicago began on June 15, 2019. During the nearly four-month
period, 2,500 e-scooters were deployed, which generated more than 800,000 trips (CDOT, 2019).
The Chicago bike sharing system began operation on June 27, 2013, with 75 stations and 750
bicycles. By December 2019, the number of stations had increased to 612, and the number of
bicycles exceeded 6,000. The usage had also increased year by year. We aim to quantitatively
analyze the impact of the e-scooter sharing on the usage of bike sharing. A difference-in-
differences (DID) model based on the propensity score matching (PSM) method is used for this
analysis.
This paper is further structured as follows. Section 2 introduces the literature review on e-
scooter sharing and bike sharing. Section 3 describes the Chicago e-scooter sharing and bike
sharing data and performs a comparative analysis. Section 4 presents the method used in this article.
Section 5 presents the analysis and discusses the results. Section 6 concludes and discusses the
limitations of this study.
2. Literature Review
In this study, we aim to explore the impact of the advent of dockless e-scooter sharing on the
usage of station-based bike sharing. In this section, we describe the literature on the role of bike
sharing and e-scooter sharing in the transportation system, followed by a review of the studies
exploring the interaction between station-based bike sharing and dockless e-scooter sharing.
Factors that influence the demand for bike sharing include the “5D” built environment, trip
characteristics, user characteristics, and weather (Faghih-Imani and Eluru, 2016; Wu et al., 2021;
Yang et al., 2020). Because the demand profiles often follow morning and afternoon peak hours,
3
commuting is regarded as one of the major trip purposes of bike sharing (Bordagaray et al., 2016;
Lathia et al., 2012; McKenzie, 2019; Zhao et al., 2015). Other scholars have done work on the
interaction between bike sharing and other modes of transportation (Campbell and Brakewood,
2017; Gu et al., 2019; Guo and He, 2021; Kong et al., 2020; Li et al., 2019; Ma et al., 2019). Three
types of interactions have been defined between bike sharing and other modes of transportation:
substitution, integration, and complementarity (Kong et al., 2020). Most often, bike sharing mainly
replaces public transit (Campbell and Brakewood, 2017; Ma et al., 2019). In these studies, the
difference in difference (DID) methods are commonly used. For example, Campbell and
Brakewood (2017) and Ma et al. (2019) used the DID approach to analyze the changes in bus
ridership before and after the emergence of bike sharing. The results show that the emergence of
bike sharing can reduce bus ridership.
Early research on e-scooter sharing relied on questionnaire-based stated preference surveys
to collect data. Those studies link demographic characteristics of shared e-scooter users and their
intention to use shared e-scooters (Aguilera-García et al., 2020; Eccarius and Lu, 2020; Laa and
Leth, 2020; Mitra and Hess, 2021; Sanders et al., 2020). Men and people with higher education
tend to be more likely to use shared e-scooters (Aguilera-García et al., 2020; Laa and Leth, 2020;
Mitra and Hess, 2021; Nikiforiadis et al., 2021).
When e-scooter sharing trip data became available, many studies analyzed the impact of the
built environment on behavior (Bai and Jiao, 2020; Bai et al., 2021; Caspi et al., 2020; Hawa et al.,
2021; Huo et al., 2021). These studies found that e-scooter sharing is most heavily used around
university campuses and in downtown areas, and deployment patterns follow this demand. The
peak hours for shared e-scooter trips are usually in the afternoon. In terms of replaced transport
modes, Laa and Leth (2020); Mitra and Hess (2021); Nikiforiadis et al. (2021) found that e-scooter
sharing mainly replaced walking and public transportation, while Sanders et al. (2020) found that
e-scooter sharing replaced many walking and bicycling trips. On the other hand, PBOT (2018)
found that a large proportion of shared e-scooter trips (around 34%) replaced car-based trips.
Some studies explored the interaction between e-scooter sharing and bike sharing and their
findings are relevant to the topic of this study. For example, Zhu et al. (2020) analyzed bike sharing
and e-scooter sharing in Singapore and found that bike sharing is often used for commuting while
e-scooter sharing mainly serves recreation or tourism activities in the downtown area. This finding
is consistent with that of McKenzie (2019) and Zamir et al. (2019), both of which are based on the
analysis of bike sharing and e-scooter sharing, trips in Washington D.C.. Reck et al. (2021)
analyzed the choice behavior of the four shared micro-mobility modes (docked e-bike, docked
bike, dockless e-scooter, dockless e-bike) and found that docked modes are more preferred for
commuting than dockless. Reck and Axhausen (2021) surveyed the residents of Zurich and found
that the demographics of e-scooter sharing users are more similar to those of the residents than
bike sharing users, which indicates that e-scooter sharing users are more representative of the
residents. Younes et al. (2020) analyzed the determinants of demand for e-scooter sharing and bike
sharing in Washington DC and the interaction between the two modes. They found that the usage
of e-scooter sharing has a possible competitive relationship with the usage of bike sharing by non-
members and a complementary relationship with the usage of bike sharing by members. However,
they acknowledge the lack of causality in their study.
As a result, to our best knowledge, no research has analyzed the impact of e-scooter sharing
on the usage of bike sharing using long-term data. Therefore, we perform this study aiming to
answer the following three questions: (1) Whether e-scooter sharing impacts the usage of bike
sharing; (2) Whether the impact is different among different bike sharing user types; (3) Whether
4
the impact is different among different types of trips (happening in different time periods or are
different durations)?
3. Data
3.1 Background
On June 15, 2019, ten companies (Bird, Bolt, Grüv, Jump, Lime, Lyft, Sherpa, Spin, VeoRide,
and Wheels) obtained permits to operate e-scooter sharing in the Chicago Department of Business
Affairs and Consumer Protection. Altogether, 2,500 e-scooters were allowed to be operated in
designated areas in the northwest of the city. The e-scooter sharing pilot program in Chicago aimed
to provide residents with a fair, safe, and sustainable travel mode while testing the performance of
e-scooter systems and providing equitable access. The northwest area was designated as the pilot
program area because of the areas many different demographic groups and a variety of
transportation modes (CDOT, 2019). Table 1 shows the number of e-scooter trips reported by 10
operating companies during the pilot (spanning four months). By comparing the trip data reported
by various operators, Bird, Lime, Lyft, and Jump have a higher market share than the other
operators. Each operator sets its own price. Generally, the price is one dollar to start the trip and
15 cents for every minute of use (CURBED, 2019).
Table 1 E-scooter sharing operators and number of trips provided
Provider
Number of trips
Proportion
Bird/Sherpa
178,134
21.7%
Lime
121,131
14.7%
Lyft
119,116
14.5%
Jump
100,528
12.2%
VeoRide
75,559
9.2%
Grüv
68,620
8.4%
Wheels
57,740
7.0%
Spin
55,463
6.8%
Bolt
45,324
5.5%
The Chicago Divvy bike sharing was first launched on June 27, 2013, and was operated by
Motivate, which was later bought by Lyft. More than 612 stations and 6,000 bikes exist now. In
2019, Divvy’s average daily ridership was approximately 10,460, and each bike is used
approximately 1.74 times per day. The membership includes three types; annual member, 15-day
member, and non-member. Non-member users pay $3 per trip, and the trip duration is limited to
30 minutes. Beyond 30 minutes, the price increases by $0.15 for every additional 1 minute. Figure
1 shows the location and number of docks of bike stations. Stations are primarily located in the
downtown area of Chicago, with less dense distributions beyond the urban core. In 2016, more
than 10 million trips used Divvy bike sharing.
5
Figure 1 Location of bike sharing stations and e-scooter sharing pilot program area
3.2 Data Description
The data used in this study include three datasets, Divvy bike sharing ridership data
1
, e-
scooter sharing ridership data
2
, and the Smart Location Database
3
. There were more than 2,870,000
bike sharing trips from March 2019 to October 2019. The trip-related information includes starting
and ending time, starting and ending station, trip duration, type of user (annual members, 15-day
members, and non-members), and age and gender of annual and 15-day members (the
demographic information of non-members is unknown). The bike sharing data also contain the
coordinates and capacity of each bike station.
E-scooter sharing ridership data are from June 15, 2019, to October 15, 2019. There were
812,615 reported shared e-scooter trips during that time period across operators (CDOT, 2019).
Only 664,975 disaggregated trips are stored and provided to the public for the time period between
June 15 and October 15. Among them, a significant portion of the trips does not have latitude or
longitude information. Those trips are deleted, leaving 560,449 trips. We further deleted the trips
shorter than 100 m, longer than 10 km, less than 60 seconds, or longer than 2 hours, leaving
557,462 trips to be analyzed (deleted 0.54% of the trips). These criteria are consistent with many
1
https://www.divvybikes.com/system-data
2
https://data.cityofchicago.org/Transportation/E-Scooter-Trips-2019-Pilot/2kfw-zvte
3
https://www.epa.gov/smartgrowth/smart-location-database-technical-documentation-and-user-guide
6
previous studies (Huo et al., 2021; Shen et al., 2018). When we further cleaned the data, we found
that the data from October 1 to October 14 were missing. As a result, we set the study period as
June 17- September 29 (15 weeks). The number of trips during the study period is 538,287.
E-scooter sharing trip data include the census tract of trip origin and destination, trip starting
and ending time, trip duration, and distance. In addition, socio-economic and built environment
covariates around each bike station were collected to control the effects of those covariates. The
built environment covariates include the “5Ds covariates (i.e., density, diversity, design, distance
to transit, and destination accessibility) (Ewing and Cervero, 2010), which were obtained from the
Chicago data portal (2018)
4
. Then, the socio-economic variables were drawn from the Smart
Location Database (Ramsey and Bell, 2014) and the US Census Bureau
5
.
3.3 Data Analysis
3.3.1 Divvy bike sharing ridership analysis
Table 2 The usage of bike sharing in different time periods
Time period (trips/hours)
Weekday or
weekend
Sum
00:00-
05:59
06:00-
10:59
11:00-
15:59
16:00-
18:59
19:00-
23:59
Week
day
Week
end
Gend
er
Male
10,393
132,889
115,268
252,815
67,845
1936,
634
464,1
86
2,400,8
20
Femal
e
2,970
45,598
46,764
86,723
23,636
641,4
14
216,5
64
857,978
User
type
Memb
er
12,282
169,359
135,962
310,267
81,253
2,393,
963
543,4
04
2,937,3
67
Non-
memb
er
3,773
22,844
72,674
78,461
29,005
502,4
02
378,2
35
880,637
Age
55-73
2,618
19,746
14,791
18,887
4,211
212,6
87
84,16
6
296,853
41-54
4,764
43,631
25,772
46,177
9,716
451,5
66
179,1
82
630,748
31-40
7,507
76,122
45,502
95,911
26,430
821,5
25
327,3
53
1,148,8
78
25-30
9,099
68,136
50,444
107,032
38,381
845,2
72
336,0
45
1,181,3
17
19-24
2,551
16,169
18,269
31,419
14,921
251,9
24
100,1
47
352,071
Table 2 lists bike sharing ridership in each period by gender, user type, and age group.
According to the temporal distribution of ridership, the time period is divided into five parts: AM
peak hours (6:0010:59), Midday (11:0015:59), PM peak hours (16:0018:59), Evening (19:00
23:59), and Night (0:005:59). We can see that trips made by male users are roughly triple those
4
https://data.cityofchicago.org/
5
https://data.census.gov/cedsci/
7
made by female users. Regarding the user type, trips made by members are more than triple those
made by non-members. The period with the highest number of trips for members is from 16:00 to
18:59, and for non-members is from 11:00 to 15:59. Regarding age, the highest number of trips is
made by the 2530 age group, followed by the 3140 age group. Notably, the highest number of
trips for users of age 1954 is in the evening peak hours from 16:00 to 18:00, whereas for users of
age 55 to 73 is from 06:00 to 10:00. This finding may indicate that users of age 1954 mainly use
bike sharing for commuting, whereas users of age 5573 mainly use bike sharing for non-
commuting activities.
3.3.2 Distribution of trip duration of bike sharing and e-scooter sharing
Figure 2 Distribution of trip duration of bike sharing and e-scooter sharing
In Figure 2, we plotted the distribution of trip duration of bike sharing and e-scooter sharing.
The average duration of bike sharing trips made by non-members is twice that of members. The
duration of e-scooter trips appears to be similar to that of bike trips made by members.
3.3.3 Temporal distribution of ridership of bike sharing and e-scooter sharing
8
Figure 3 Number of bike sharing trips by hour
Figure 4 Number of e-scooter sharing trips by hour
9
As expected, the temporal distribution of bike sharing ridership is different on weekdays and
weekends (Figures 3 and 4). On weekdays, the ridership has two peak periods, that is, morning
and afternoon peak hours, indicating that that bike sharing is an important transportation mode for
commuting, particularly for members. From a different perspective, the ridership of e-scooter
sharing on weekdays and weekends are similar, both without the two-peak weekday commute
pattern. The peak occurs late in the afternoon, between 16:00 and 20:00 on weekdays and 14:00
and 19:00 on weekends. The weekday peak includes typical afternoon commute hours.
4. Research Design and Method
4.1 Method
Difference-in-difference (DID) is a widely used statistical method to infer a causal
relationship in social science to mimic an experimental research design using observational data
(Ashenfelter and Card, 1984; Li et al., 2018; Li et al., 2012; Li et al., 2019), particularly when a
randomized experiment is impossible as in our case. The basic premise is to divide all the studied
units into treated and untreated units. The former receives treatment, which is the influence of the
e-scooter sharing in our study, whereas the latter does not receive treatment. The effect of the
treatment is estimated by comparing the change of the outcome (usage of bike sharing in this study)
over time (before and after the launching of e-scooter sharing) of the treated units to the change of
the outcome of the untreated ones. This could mitigate the effects of extraneous factors (e.g.,
systematic changes in demand).
However, the results of the DID method could suffer from selection bias, which means that
the units in the treated group are inherently different from those in the untreated group. This would
make the treated and untreated units not comparable, and the results of the DID analysis unreliable.
In Propensity Score Matching (PSM), the propensity score refers to the probability that a unit with
certain characteristics belongs to the treated or untreated units. The scores can be used to reduce
or eliminate potential selection bias by balancing the characteristics of the units between the treated
and untreated ones. PSM creates matched sets of treatment and control groups in the treated and
untreated units. A matched set consists of at least one unit in the treatment group and one unit in
the control group with a similar propensity score. This method could ensure that the outcome is
independent of the division of the treatment and control groups. After matching, the difference in
the outcomes observed between the treatment and control groups can be completely attributed to
the treatment (Rosenbaum and Rubin, 1983).
4.2 Theoretical framework for DID-PSM
The psmatch2 package of the Stata software is used to implement the DID-PSM (Leuven
and Sianesi, 2003). The following sections describe the steps to implement DID-PSM.
(1) Propensity score estimation
Given that the propensity score indicates the probability that a unit belongs to the treated units,
logit and probit models are usually used to construct the relationship between the covariates
(characteristics of the units) and the group category. In this study, the logit model is selected:


where X represents the covariates, T represents the group category with 1 indicating treated units
and 0 indicating untreated units, is the intercept, and is the vector of regression coefficients.
(2) Matching algorithm and balance detection
10
After obtaining the propensity score, we need to use matching algorithms to select
observations from the untreated units to construct the control group. Common matching algorithms
include k-nearest neighbor, caliper matching, radius matching, and kernel matching. To date, no
conclusion exists on which is the best matching algorithm. Generally, when the sample size is
large, the results of all the matching algorithms should be similar (Li et al., 2013). More details on
the matching algorithms can be found in (Heinrich et al., 2010). Usually, multiple matching
algorithms are used and compared to increase the reliability of the results. In this study, we use
balance detection to test the reliability of matching results. If no significant difference exists in the
covariates used for matching between the two groups, then the matching covariates and matching
methods selected are suitable, and the matching result is good. We use the following equation to
calculate the standardized difference : 
where:  and represent the mean and standard deviation of a covariate of the treatment group,
respectively;  and represent the mean and standard deviation of a covariate of the control
group. Some studies suggest that when the difference  for all covariates, it passes the
balance test and the result is acceptable (Wang and Cao, 2017).
(3) Treatment effect estimation
After the balance test is passed, the average difference between the treatment and control
groups is reflected by the average treatment effect (ATE). Considering that the t-value of ATE is
not provided in Stata, we choose the average treatment effect on the treated (ATT) to represent the
change of bike sharing usage after the launching of e-scooter sharing. The equation of ATT is as
follows: 



where and represent the treatment and control groups, respectively;  represents the
difference in the usage of shared bikes before and after the treatment for the  station;
represents the weight of the station, which is obtained by various matching methods.
4.3 Treated units and study period
We study the impact of e-scooter sharing on the use of bike sharing using the DID model
based on the PSM method. Hence, we first need to divide the bike sharing stations into treatment
and control units. Treated units refer to the stations for which the usage of bike sharing is
influenced by the e-scooter sharing, whereas the untreated units refer to the contrary.
We regard the bike sharing stations located in the e-scooter sharing pilot program with no
less than 100 shared e-scooter trips within the 400-meter buffer area around the station as the
treated units. Since only the census tract of the shared e-scooter trip origin is known, we regard
the shared e-scooter trips to be uniformly distributed in the census tract so as to calculate the
number of shared e-scooter trips within the buffer area. The reason for choosing the 400-meter
buffer area is that previous studies have concluded that 400 meters is considered an acceptable
walking distance to access the bike sharing station (Daniels and Mulley, 2013; Hoehner et al.,
11
2005; Li et al., 2019; McCormack et al., 2008; Pangbourne, 2019; Pikora et al., 2003). The number
of units in the treated and untreated units is 150 and 461, respectively, with a ratio of approximately
1:3, which is considered sufficient to ensure the quality of the matching.
Figure 6 shows the location of bike sharing stations and spatial variation of e-scooter trips.
The red dots represent stations in the treated units, whereas the blue ones represent stations in the
untreated units. The figure also shows the number of e-scooter trips of each census tract.
The outcome variable is the difference in the change in usage before and after the launching
of e-scooter sharing. To obtain a reliable estimation of the effect of the treatment, we select the
time period of 15 weeks before and after the launching of e-scooter sharing, from March 2, 2019,
to September 29, 2019. The period of 15 weeks before the launching of e-scooter sharing is
regarded as the pre-treatment period, and that after the launching of e-scooter sharing is regarded
as the post-treatment one.
Figure 6 Location of bike sharing stations and spatial variation of e-scooter trips
4.4 Covariates
The PSM method requires covariates to capture as many potential confounding factors as
possible. Therefore, the model should include covariates that affect bike sharing trips. However,
adding a covariate that has no impact on the outcome in PSM would increase the variance of the
estimated treatment effect (Brookhart et al., 2006). Thus, all the covariates that were associated
12
with outcomes should be included regardless of whether they influence the treatment assignment
(Brookhart et al., 2006; Li et al., 2017).
The selection of covariates in our study is based on previous studies that explored factors
affecting bike sharing usage (Faghih-Imani and Eluru, 2016; Faghih-Imani et al., 2014; Fishman
et al., 2014; Li et al., 2018; Li et al., 2019). Previous studies have shown that socio-economic
covariates, such as population density, income, and employment density, affect the usage of bike
sharing (Faghih-Imani and Eluru, 2016; Wang et al., 2018; Yang et al., 2020). Other studies found
that land use factors and transportation infrastructure-related covariates also influenced the usage
of bike sharing (Yang et al., 2020). Therefore, we choose to include three types of covariates:
socio-economic, land use, and transportation infrastructure. A buffer with a radius of 400 meters
is drawn around each station. The reason for choosing 400 meters is as follows. First, Daniels and
Mulley (2013); Noland et al. (2016) recommended that 400 meters is generally considered as the
distance threshold for residents to walk to public transportation and bicycle stations. Second, when
the distance is too low, the surrounding variable attributes may not be captured since the resolution
of the census tracts is larger than very short walking distances. The value of each variable is
extracted from the buffer area. Given that the boundary of the census block group (CBG) is not
the same as the buffer, residents and jobs are assumed to be uniformly distributed in the CBG.
Thus, the values of these covariates for each station could be calculated based on this assumption.
Based on previous studies, the following covariates are selected (Table 3). Socioeconomic
covariates include population density, employment density, and average income. Land use
covariates include residential area, commercial area, industrial area, park area, and distance to the
city center. Transportation infrastructure covariates include bike trail density, major road density,
number of bus stops, passenger volume of nearby subway stations within the 400-m buffer, number
of docks of nearby bike stations within the 400-m buffer, and traffic volume (average daily traffic
volume of the road closest to the station, or AADT) of the road closest to the bike station (Zhang
et al., 2017).
Table 3 Descriptive statistics of covariates to calculate propensity score
Covariates
Description
Mean
Min
Max
Std
Bike trips
Weekly bike sharing trips
130
1.000
4,172
206
E-scooter trips
Weekly e-scooter sharing
trips
4,202
51,946
68,611
19,854
Population
density
Population density within
the 400m buffer
4,240
312
18,500
2,637
Employment
density
Employment density within
the 400m buffer
8,657
11.637
175,768
24,798
Average income
Average annual income of
residents within the 400m
buffer (dollar)
79,807
2,223
834,629
93,878
Bike trail density
Density of bike trail within
the 400m buffer (km/km2)
1.648
0.000
6.477
1.458
Major road
density
Density of major road
within the 400m buffer
(km/km2)
0.043
0.000
0.499
0.064
Bus stop
Number of bus stops within
the 400m buffer the 400m
buffer
8.913
0.000
27.000
4.970
13
Subway ridership
Number of inbound
passengers of subway
stations within the 400m
buffer
535,794
0.000
16,916,972
1,614,785
Number of docks
Total number of docks of
stations within the 400m
buffer
27.601
7.000
170.000
28.846
Residential area
Proportion of residential
area within the 400m buffer
0.290
0.000
0.633
0.167
Commercial area
Proportion of commercial
area within the 400m buffer
0.130
0.000
0.802
0.120
Industrial area
Proportion of industrial
area within the 400m buffer
0.031
0.000
0.543
0.075
Park area
Proportion of park area
within the 400m buffer
0.062
0.000
0.954
0.141
AADT
Average daily traffic
volume of the road closest
to the station
10,111
154.075
28,700
6,102
Distance to city
center
Distance from the station to
city center (m)
4,240
312
18,500
2,637
5. Results
5.1 Initial analysis
In this section, we perform an initial analysis of the usage of bike sharing in Chicago. Figure
7 shows the average weekly number of bike sharing trips in the treated and untreated units over
time. The average usage of the untreated stations was lower than that of the treated stations before
e-scooter sharing was introduced and became higher than that of the treated units after e-scooter
sharing was introduced.
14
Figure 7 Bike sharing average weekly usage before and after the introduction of e-
scooter sharing.
5.2 PSM quality evaluation
It is crucial to make sure the units in the control group have similar characteristics to the units
in the treatment group. The overlap test is a routine test that is performed to evaluate whether the
units in the two groups are similar. The validity of the PSM results is evaluated by visually
inspecting the distribution of propensity scores of units in the treatment and control groups, which
is shown in Figure 8. If there are both treated units and untreated units in a column, those units are
called “on support.” Otherwise, those units are called “off support” and should be deleted before
further analysis (Li et al., 2018). Figure 8 shows that all the units are “on support” and could be
used for further analysis.
15
Figure 8 Results of overlap test based on propensity score
A balancing test is performed to check whether the stations in the treatment and control
groups are similar after matching, which could mean that no significant difference exists in the
mean values of covariates between the stations in the treatment and control groups. Table 4 shows
the t-test results of the difference in the mean values of covariates before and after matching. The
results show that significant differences exist in most of the mean values of covariates between
units of the treatment group and groups before matching (p < 0.05). After matching, no significant
difference is observed.
Table 4 Balancing test results
Variable
Unmatched or
matched
Mean of
treatment
group
Mean of
control
group
% bias
% |bias|
reduced
t-stat
p value
Population
density
U
4,209.700
4,429.300
-9.400
-2.290
0.022
M
4,209.700
4,127
3.600
62.300
0.930
0.353
Employment
density
U
2,184.700
10,634
-42.500
-9.120
0.000
M
2,184.700
2,091.700
0.500
98.900
0.740
0.458
Mean income
U
73,461
87,267
-17.100
-3.980
0.000
M
73,461
74,492
-1.300
92.500
-0.390
0.695
Bike trail
density
U
1.805
1.597
15.100
3.990
0.000
M
1.805
1.770
2.500
83.200
0.550
0.585
Major road
density
U
0.038
0.041
-5.200
-1.550
0.120
M
0.038
0.038
-0.200
96.300
-0.040
0.964
16
Bus stop
U
7.377
9.135
-37.600
-9.840
0.000
M
7.377
7.449
-1.500
95.900
-0.360
0.717
Subway
passengers
U
200,000
630,000
32.100
-7.070
0.000
M
200,000
220,000
-1.100
96.600
-0.560
0.577
Number of
docks
U
20.164
29.758
-40.000
-9.060
0.000
M
20.164
20.746
-2.400
93.900
-0.980
0.326
Residential
area
U
0.339
0.291
30.200
8.140
0.000
M
0.339
0.333
3.600
88.100
0.770
0.444
Commercial
area
U
0.126
0.133
-5.800
-1.520
0.128
M
0.126
0.132
-4.900
15.600
-1.060
0.289
Industrial area
U
0.038
0.022
23.400
6.520
0.000
M
0.038
0.039
-2.700
88.300
-0.480
0.629
Park area
U
0.032
0.072
-31.200
-7.680
0.000
M
0.032
0.033
-1.000
96.300
-0.310
0.759
AADT
U
11,946
11,080
13.500
3.800
0.000
M
11,946
11,444
7.800
42.100
1.610
0.108
Distance to
center
U
7,803
10,660
-56.000
-12.900
0.000
M
7,803
7,725
1.500
97.300
0.500
0.616
5.3 Impact of e-scooter sharing on the usage of bike sharing
5.3.1 Overall impact
In this section, we evaluate the impact of e-scooter sharing on the usage of bike sharing. When
using the PSM method for analysis, a variety of matching methods are used to verify the validity
of the results. In our study, the matching methods used include K-nearest neighbor, kernel, and
radius matching. Table 5 shows the results of those matching methods. After e-scooter sharing is
introduced, the average weekly usage of bike sharing is reduced substantially (by 10.2%).
Moreover, the matching results of different methods are similar, ranging from 9.7% to 11.3%. In
the following analysis, we report the results of the kernel matching analysis.
Table 5 The impact of e-scooter sharing on weekly bike sharing usage
Models
Sample
Treatme
nt Group
Control
Group
Differenc
e
t-stat
K-nearest neighbors matching (K = 3)
PSM
model
68.197
90.759
-22.562
-4.970*
K-nearest neighbors matching (K = 5)
PSM
model
68.197
90.440
-22.243
-5.160*
Radius matching (caliper = 0.01)
PSM
model
68.309
90.603
-22.295
-3.780*
Radius matching (caliper = 0.1)
PSM
model
68.197
95.054
-26.858
-5.790*
Kernel (bandwidth=0.05)
PSM
model
68.197
92.081
-23.884
-5.240*
* |t-stat| >1.96
5.3.2 Impact on the usage of bike sharing by type
17
Table 6 Effect of e-scooter sharing on different user types of bike sharing
Type
Group
Sample
Treatment
Group
Control
Group
Difference
t-stat
Effect
Members
hip
Member
PSM
model
48.193
53.544
-5.350
-2.316*
-3.95%
Non-member
PSM
model
20.206
38.235
-18.029
-6.572*
-34.0%
Member-
Gender
Male
PSM
model
30.576
33.897
-3.321
-2.095*
-3.1%
Female
PSM
model
17.569
19.548
-1.979
-2.246*
-7.3%
Member-
Age
19-24
PSM
model
9.569
11.759
-2.190
-3.520*
11.4%
25-30
PSM
model
19.358
21.058
-1.700
-2.695*
3.3%
31-40
PSM
model
12.128
12.310
-0.182
-0.423
41-54
PSM
model
3.398
3.9271
-0.529
-0.491
55-73
PSM
model
3.740
4.490
-0.750
-0.300
Trip
duration
0-15 min
PSM
model
36.204
49.119
-12.915
-6.070*
-15.8%
15-30 min
PSM
model
23.388
28.826
-5.438
-2.890*
-21.2%
30-60 min
PSM
model
5.164
8.592
-3.428
-2.810*
-16.1%
Different
time
periods
Weekday peak
hours
PSM
model
38.311
42.892
-4.581
-1.830
Weekday non-
peak hours
PSM
model
15.289
23.549
-8.261
-4.940*
-7.5%
Weekend peak
hours
PSM
model
7.182
13.984
-6.802
-5.340*
-9.6%
Weekend non-
peak hours
PSM
model
7.414
11.404
-3.99
-4.090*
-20.5%
Table 6 shows the impact of e-scooter sharing on the usage of bike sharing by the member
type (member and non-member). In general, after e-scooter sharing is introduced, the usage of bike
sharing by members and non-members decreased by 5.4 (3.95%) and 18.0 (34.0%) trips per week
per station, respectively. e-scooter sharing had a much larger impact on non-member ridership.
The age and gender of non-member bike sharing riders are unknown. However, in this study, we
focus on the impact of e-scooter sharing on the usage of bike sharing members by gender and age
group. Table 6 shows the influence of e-scooter sharing on the usage of bike sharing members by
gender. The results show that after e-scooter sharing was introduced, ridership of male and female
members decreased by 3.3 and 2.0 trips per week, respectively. We categorize age into five groups:
18
19-24, 25-30, 31-40, 41-54, and 55-73, which corresponds with the age range usually used in
previous studies (Lee et al., 2019; Newbold and Scott, 2017, 2018; Rahimi et al., 2020; Wang et
al., 2018; Wang, 2019). Age groups may be as important, or more important, than other
generational factors. Table 6 shows the impact of e-scooter sharing on bike sharing usage of
members of different age groups. The results show that after the introduction of e-scooter sharing,
the bike sharing usage of the age group 19-30 decreases significantly. Other age groups did not
see large or significant changes. This is probably because e-scooter sharing is more attractive to
young users (age 19-31) than others (Mitra and Hess, 2021; Shaheen et al., 2013).
Based on the temporal variation of bike sharing trips, on weekdays, the morning and evening
peak hours are 6:0010:00 and 16:0019:00, respectively. On weekends, there is a relatively
uniform peak from 10:00 to 18:00. The results show that the introduction of e-scooter sharing does
not have a significant impact on the use of bike sharing during peak hours on weekdays. However,
we saw reductions in trips per week during non-peak hours on weekdays (8.3 (7.5%)) and peak
(6.8 (9.6%)) and non-peak hours (4.0 (20.5%)) on weekends. Thus, only trips of bike sharing
during commute hours are not affected by the introduction of e-scooter sharing.
Non-member bike sharing is priced at 3 dollars for a 30-minute trip and 0.15 dollars for every
additional minute. Therefore, we use 30 minutes as the division interval. We divide bike sharing
trips into three groups based on the trip duration: short-duration trips (015 min), medium-duration
trips (1530 min), and long-duration trips (3060 min). In this study, we find that the number of
short-, middle-, and long-duration trips decrease by 12.9 (15.8%), 5.4 (21.2%), and 3.4 (16.1%),
respectively.
6. Discussion
Our results show that, on average, the weekly usage of bike sharing decreases by 10.2%. The
weekly usage of bike sharing of members decreases by 4.0% and that of non-members decreases
by 34.0%. Younes et al. (2020) concluded that e-scooter sharing had a complementary relationship
with bike sharing for bike sharing members and a competitive relationship with bike sharing for
bike sharing non-members. Our finding refutes the first part of the finding of Younes et al. (2020).
We found that e-scooter sharing also has a competitive relationship with bike sharing members
and the competition is more intense for bike sharing non-members.
Because only the gender and age of members are known, we can only study the impact of the
advent of e-scooter sharing on the usage of members of different genders and age groups. All of
the changes of bike sharing ridership come from younger riders; ridership of members of the 19-
41 age group decreased by 3.4%, while ridership of other age groups was not significantly
impacted. This result is consistent with the finding of previous studies that e-scooter sharing users
are more likely to be young (Huo et al., 2021; Mitra and Hess, 2021; Shaheen et al., 2013). The
bike sharing ridership of male and female members decreased by 3.1% and 7.3%, respectively.
Short, medium, and long duration bike sharing trips decreased by 15.8%, 21.2%, and 16.1%,
respectively. This contrasts with the effects of dockless bike sharing on station-based bike sharing
shown by (Li et al., 2019), which shows that only short duration trips decreased, while the ridership
of middle and long duration trips was not significantly influenced.
We compared the average daily trips of e-scooter sharing and bike sharing on weekdays and
weekends and found that the peak hours for shared e-scooter trips on weekdays or weekends are
about the same, late afternoon (16:0018:00), whereas the peak hours for bike sharing trips are
6:009:00 and 17:0018:00 on weekdays (reflecting commute-oriented trips) and 11:0016:00 on
weekends. We also observed that the introduction of e-scooter sharing does not have a significant
19
impact on the usage of bike sharing during peak hours on weekdays. But for non-peak hours on
weekdays, peak hours on weekends, and non-peak hours on weekends, the bike sharing usage
decreases by 7.5%, 9.6%, and 20.5%, respectively. Again, e-scooter sharing has the largest impact
on the weekend or non-commute-oriented trip patterns. This likely reflects the larger marginal cost
of use from e-scooter sharing that is more akin to recreation or visitor bike sharing trips, compared
to bike sharing subscription rates that result in a very low marginal cost that caters to repeat or
habitual commute patterns. This finding is consistent with the finding of previous studies that
commuting is not the major trip purpose of e-scooter sharing (Caspi et al., 2020; Reck et al., 2021).
7. Conclusions
e-scooter sharing is newly developed shared micromobility travel mode. its advent could
possibly influence the ridership of bike sharing. In this study, we try to answer the question of
whether the advent of e-scooter sharing influences the ridership of bike sharing and to quantify the
influence on the bike sharing ridership of different user groups and types of trips. The e-scooter
sharing and bike sharing trip data of Chicago are used in this study. The PSM-DID method is
adopted to rule out the effect of exogenous factors on the study results. Results show that the bike
sharing ridership declines by 10.2% in the e-scooter sharing operation area due to the advent of e-
scooter sharing. Bike sharing operators should anticipate a decrease in the usage of bike sharing
when e-scooter sharing is introduced and lower the expectation of ridership and revenue;
specifically, non-commute-oriented trips and trips with a higher marginal revenue stream will
likely erode. Based on the quantification of the impact of e-scooter sharing on bike sharing usage,
the bike sharing operators could adjust the allocation of resources for operation and maintenance
of bike sharing, as well as the pricing scheme to maximize the utility of bike sharing. But the bike
sharing in Austin could still continue to operate because 90% of the existing ridership remains.
The ridership of bike sharing members and non-members decreases by 4.0% and 34.0%
respectively. When e-scooter sharing and bike sharing are both under operation in the same area,
the locations and capacities of the stations of bike sharing could be adjusted to focus more on the
member users. The bike sharing ridership by members of the age group 19-40 decreases by 3.4%
while that of the other age groups does not decrease significantly. Regarding the ridership of
different time periods, the ridership during weekday non-peak hours, weekend peak hours, and
weekend non-peak hours decreases by 7.5%, 9.6%, and 20.5%, respectively, while the ridership
during weekday peak hours does not change significantly. Commuting is not one of the major
purposes of shared e-scooter trips (Caspi et al., 2020; Huo et al., 2021; Noland, 2019). These results
reveal that the decrease of bike sharing ridership is not uniformly distributed among different user
groups or types of trips. The locations and capacities of bike sharing should be adjusted to
accommodate for the change caused by the advent of e-scooter sharing.
This study also has some limitations. First, this study is based on the Chicago case under the
specific local settings of the e-scooter sharing and bike sharing system and specific characteristics
of Chicago’s e-scooter sharing trial. Chicago’s e-scooter sharing program specifically aimed to
reduce overlap between the systems, so it is possible that other less regulated systems may
experience more competition. The result could be influenced by characteristics of the e-scooter
sharing system, such as the price of using shared e-scooters, how large the e-scooter sharing
operation area is, and the number of e-scooters that are placed in the operation area. With a lower
price of using e-shared e-scooters, larger operation area, and higher number of e-scooters, one may
observe larger decreases of bike sharing ridership. However, bike sharing systems that have
responded to e-scooter sharing by changing their pricing structure to be more competitive with the
20
price of using shared e-scooters could stem some of the competitive impacts of ridership
(Venigalla et al., 2020). Secondly, whether the results of this study could be applied to other cities
is not clear. Yet, our results are clear. In this context, e-scooter sharing has a competitive
relationship with the bike sharing system instead of a complementary relationship. E-scooter
sharing significantly and substantially eroded ridership of the bike sharing system, particularly
among the non-members. Bike sharing system operators should anticipate a decline in bike sharing
ridership and revenue when the e-scooter sharing is implemented in the city and should adjust the
locations and capacities of the bike sharing stations and pricing to accommodate for this change.
In this study, we propose a framework to analyze the impact of e-scooter sharing on the usage of
bike sharing. Future research could use this framework to perform similar analyses based on data
from other cities to obtain more insights.
Acknowledgments
This study was funded by the National Natural Science Foundation of China (grant number
71704145, 51774241, and 71831006), China Postdoctoral Science Foundation, and Sichuan Youth
Science and Technology Innovation Research Team Project (grant number 2019JDTD0002 and
2020JDTD0027).
21
References
Aguilera-García, Á., Gomez, J., Sobrino, N. (2020) Exploring the adoption of moped scooter-
sharing systems in Spanish urban areas. Cities 96, 102424.
National Bureau of Economic Research (1984) Using the longitudinal structure of earnings to
estimate the effect of training programs.
Austin (2019) What 5 Million Scooter Rides in Austin Tell Us About the Future of Micromobility.
Bai, S., Jiao, J. (2020) Dockless E-scooter usage patterns and urban built environments: a
comparison study of Austin, TX, and Minneapolis, MN. Travel behaviour and society 20, 264-
272.
Bai, S., Jiao, J., Chen, Y., Guo, J. (2021) The relationship between E-scooter travels and daily
leisure activities in Austin, Texas. Transportation Research Part D: Transport and Environment
95, 102844.
Bordagaray, M., dell’Olio, L., Fonzone, A., Ibeas, Á. (2016) Capturing the conditions that
introduce systematic variation in bike-sharing travel behavior using data mining techniques.
Transportation Research Part C: Emerging Technologies 71, 231-248.
Brookhart, M.A., Schneeweiss, S., Rothman, K.J., Glynn, R.J., Avorn, J., Stürmer, T. (2006)
Variable selection for propensity score models. American journal of epidemiology 163, 1149-1156.
Campbell, K.B., Brakewood, C. (2017) Sharing riders: How bikesharing impacts bus ridership in
New York City. Transportation Research Part A: Policy and Practice 100, 264-282.
Caspi, O., Smart, M.J., Noland, R.B. (2020) Spatial associations of dockless shared e-scooter usage.
Transportation Research Part D: Transport and Environment 86, 102396.
(2019) E-Scooter Pilot Evaluation,
https://www.chicago.gov/content/dam/city/depts/cdot/Misc/EScooters/E-
Scooter_Pilot_Evaluation_2.17.20.pdf.
CURBED (2019) Everything you need to know about electric scooters in Chicago.
Daniels, R., Mulley, C. (2013) Explaining walking distance to public transport: The dominance of
public transport supply. Journal of Transport and Land Use 6, 5-20.
Eccarius, T., Lu, C.-C. (2020) Adoption intentions for micro-mobilityInsights from electric
scooter sharing in Taiwan. Transportation Research Part D: Transport and Environment 84,
102327.
Ewing, R., Cervero, R. (2010) Travel and the built environment: A meta-analysis. Journal of the
American planning association 76, 265-294.
Faghih-Imani, A., Eluru, N. (2016) Incorporating the impact of spatio-temporal interactions on
bicycle sharing system demand: A case study of New York CitiBike system. Journal of Transport
Geography 54, 218-227.
Faghih-Imani, A., Eluru, N., El-Geneidy, A.M., Rabbat, M., Haq, U. (2014) How land-use and
urban form impact bicycle flows: evidence from the bicycle-sharing system (BIXI) in Montreal.
Journal of Transport Geography 41, 306-314.
Fishman, E., Washington, S., Haworth, N., Mazzei, A. (2014) Barriers to bikesharing: an analysis
from Melbourne and Brisbane. Journal of Transport Geography 41, 325-337.
Gu, T., Kim, I., Currie, G. (2019) Measuring immediate impacts of a new mass transit system on
an existing bike-share system in China. Transportation research part A: policy and practice 124,
20-39.
22
Guo, Y., He, S.Y. (2021) The role of objective and perceived built environments in affecting
dockless bike-sharing as a feeder mode choice of metro commuting. Transportation Research Part
A: Policy and Practice 149, 377-396.
Hawa, L., Cui, B., Sun, L., El-Geneidy, A. (2021) Scoot over: Determinants of shared electric
scooter presence in Washington D.C. Case Studies on Transport Policy.
Heinrich, C., Maffioli, A., Vazquez, G. (2010) A primer for applying propensity-score matching.
Inter-American Development Bank.
Hoehner, C.M., Brennan Ramirez, L.K., Elliott, M.B., Handy, S.L., Brownson, R.C. (2005)
Perceived and objective environmental measures and physical activity among urban adults.
American Journal of Preventive Medicine 28, 105-116.
Huo, J., Yang, H., Li, C., Zheng, R., Yang, L., Wen, Y. (2021) Influence of the built environment
on E-scooter sharing ridership: A tale of five cities. Journal of Transport Geography 93, 103084.
Kong, H., Jin, S.T., Sui, D.Z. (2020) Deciphering the relationship between bikesharing and public
transit: Modal substitution, integration, and complementation. Transportation Research Part D:
Transport and Environment 85, 102392.
Laa, B., Leth, U. (2020) Survey of E-scooter users in Vienna: Who they are and how they ride.
Journal of Transport Geography 89, 102874.
Lathia, N., Ahmed, S., Capra, L. (2012) Measuring the impact of opening the London shared
bicycle scheme to casual users. Transportation Research Part C: Emerging Technologies 22, 88-
102.
Lee, Y., Circella, G., Mokhtarian, P.L., Guhathakurta, S. (2019) Heterogeneous residential
preferences among millennials and members of generation X in California: A latent-class approach.
Transportation Research Part D: Transport and Environment 76, 289-304.
Leuven, E., Sianesi, B. (2003) PSMATCH2: Stata module to perform full Mahalanobis and
propensity score matching, common support graphing, and covariate imbalance testing.
Li, H., Ding, H., Ren, G., Xu, C. (2018) Effects of the London Cycle Superhighways on the usage
of the London Cycle Hire. Transportation research part A: policy and practice 111, 304-315.
Li, H., Graham, D.J., Liu, P. (2017) Safety effects of the London cycle superhighways on cycle
collisions. Accident Analysis & Prevention 99, 90-101.
Li, H., Graham, D.J., Majumdar, A. (2012) The effects of congestion charging on road traffic
casualties: A causal analysis using difference-in-difference estimation. Accident Analysis &
Prevention 49, 366-377.
Li, H., Graham, D.J., Majumdar, A. (2013) The impacts of speed cameras on road accidents: An
application of propensity score matching methods. Accident Analysis & Prevention 60, 148-157.
Li, H., Zhang, Y., Ding, H., Ren, G. (2019) Effects of dockless bike-sharing systems on the usage
of the London Cycle Hire. Transportation Research Part A: Policy and Practice 130, 398-411.
Lime (2019a) Nearly 1/4 Million Parisians Are Using Dockless Electric Scooters, Study Finds.
Lime (2019b) Los Angeles First In US To Break 3 Million Electric Scooter Rides.
Ma, X., Zhang, X., Li, X., Wang, X., Zhao, X. (2019) Impacts of free-floating bikesharing system
on public transit ridership. Transportation Research Part D: Transport and Environment 76, 100-
110.
McCormack, G.R., Giles-Corti, B., Bulsara, M. (2008) The relationship between destination
proximity, destination mix and physical activity behaviors. Preventive Medicine 46, 33-40.
McKenzie, G. (2019) Spatiotemporal comparative analysis of scooter-share and bike-share usage
patterns in Washington, DC. Journal of transport geography 78, 19-28.
23
Mitra, R., Hess, P.M. (2021) Who are the potential users of shared e-scooters? An examination of
socio-demographic, attitudinal and environmental factors. Travel Behaviour and Society 23, 100-
107.
NACTO (2020) Shared Micromobility in the U.S.: 2020. National Association of City
Transportation Officials.
Newbold, K.B., Scott, D.M. (2017) Driving over the life course: The automobility of Canada’s
Millennial, Generation X, Baby Boomer and Greatest Generations. Travel Behaviour and Society
6, 57-63.
Newbold, K.B., Scott, D.M. (2018) Insights into public transit use by Millennials: The Canadian
experience. Travel Behaviour and Society 11, 62-68.
Nikiforiadis, A., Paschalidis, E., Stamatiadis, N., Raptopoulou, A., Kostareli, A., Basbas, S. (2021)
Analysis of attitudes and engagement of shared e-scooter users. Transportation Research Part D:
Transport and Environment 94, 102790.
Noland, R.B. (2019) Trip patterns and revenue of shared e-scooters in Louisville, Kentucky.
Findings, 7747.
Noland, R.B., Smart, M.J., Guo, Z. (2016) Bikeshare trip generation in New York City.
Transportation Research Part A: Policy and Practice 94, 164-181.
Pangbourne, K. (2019) Disrupting Mobility: Impacts of Sharing Economy and Innovative
Transportation in Cities. Lecture Notes in Mobility, G. Meyer, S. Shaheen (Eds.), Springer, Cham
(2017). Elsevier.
Pikora, T., Giles-Corti, B., Bull, F., Jamrozik, K., Donovan, R. (2003) Developing a framework
for assessment of the environmental determinants of walking and cycling. Social Science &
Medicine 56, 1693-1703.
Rahimi, A., Azimi, G., Jin, X. (2020) Investigating generational disparities in attitudes toward
automated vehicles and other mobility options. Transportation Research Part C: Emerging
Technologies 121, 102836.
Ramsey, K., Bell, A. (2014) Smart location database. Washington, DC.
Reck, D.J., Axhausen, K.W. (2021) Who uses shared micro-mobility services? Empirical evidence
from Zurich, Switzerland. Transportation Research Part D: Transport and Environment 94,
102803.
Reck, D.J., Haitao, H., Guidon, S., Axhausen, K.W. (2021) Explaining shared micromobility usage,
competition and mode choice by modelling empirical data from Zurich, Switzerland.
Transportation Research Part C: Emerging Technologies 124, 102947.
Rosenbaum, P.R., Rubin, D.B. (1983) The central role of the propensity score in observational
studies for causal effects. Biometrika 70, 41-55.
Sanders, R.L., Branion-Calles, M., Nelson, T.A. (2020) To scoot or not to scoot: Findings from a
recent survey about the benefits and barriers of using E-scooters for riders and non-riders.
Transportation Research Part A: Policy and Practice 139, 217-227.
Shaheen, S.A., Cohen, A.P., Martin, E.W. (2013) Public bikesharing in North America: Early
operator understanding and emerging trends. Transportation research record 2387, 83-92.
Shen, Y., Zhang, X., Zhao, J. (2018) Understanding the usage of dockless bike sharing in
Singapore. International Journal of Sustainable Transportation 12, 686-700.
Tuncer, S., Laurier, E., Brown, B., Licoppe, C. (2020) Notes on the practices and appearances of
e-scooter users in public space. Journal of Transport Geography 85, 102702.
Venigalla, M., Rayaprolu, S., Brennan, T., Kaviti, S. (2020) Could Value-based Pricing Improve
Economic Sustainability of Bikeshare? Journal of Modern Mobility Systems 1, 131-137.
24
Wang, D., Cao, X. (2017) Impacts of the built environment on activity-travel behavior: Are there
differences between public and private housing residents in Hong Kong? Transportation Research
Part A: Policy and Practice 103, 25-35.
Wang, K., Akar, G., Chen, Y.-J. (2018) Bike sharing differences among millennials, Gen Xers,
and baby boomers: Lessons learnt from New York City’s bike share. Transportation research part
A: policy and practice 116, 1-14.
Wang, X. (2019) Has the relationship between urban and suburban automobile travel changed
across generations? Comparing Millennials and Generation Xers in the United States.
Transportation Research Part A: Policy and Practice 129, 107-122.
Wu, C., Kim, I., Chung, H. (2021) The effects of built environment spatial variation on bike-
sharing usage: A case study of Suzhou, China. Cities 110, 103063.
Yang, H., Zhang, Y., Zhong, L., Zhang, X., Ling, Z. (2020) Exploring spatial variation of bike
sharing trip production and attraction: A study based on Chicago’s Divvy system. Applied
geography 115, 102130.
Younes, H., Zou, Z., Wu, J., Baiocchi, G. (2020) Comparing the Temporal Determinants of
Dockless Scooter-share and Station-based Bike-share in Washington, DC. Transportation
Research Part A: Policy and Practice 134, 308-320.
Zamir, K.R., Bondarenko, I., Nasri, A., Brodie, S., Lucas, K. (2019) Comparative Analysis of User
Behavior of Dock-Based vs. Dockless Bikeshare and Scootershare in Washington, DC. arXiv
preprint arXiv:1907.11526.
Zhang, Y., Thomas, T., Brussel, M., Van Maarseveen, M. (2017) Exploring the impact of built
environment factors on the use of public bikes at bike stations: case study in Zhongshan, China.
Journal of transport geography 58, 59-70.
Zhao, J., Wang, J., Deng, W. (2015) Exploring bikesharing travel time and trip chain by gender
and day of the week. Transportation Research Part C: Emerging Technologies 58, 251-264.
Zhu, R., Zhang, X., Kondor, D., Santi, P., Ratti, C. (2020) Understanding spatio-temporal
heterogeneity of bike-sharing and scooter-sharing mobility. Computers, Environment and Urban
Systems 81, 101483.
... In the absence of the latter three, the OD location and time information are generally used and found to be sufficient to derive them. In rare cases, additional information may also be provided with the trip data, such as the nearest metro station [25], rider type (member, single ride, and day pass) [26], membership type (annual members, 15 day members, and non-members) [27], vehicle type (i.e., bicycle or scooter) [21,26], age, and gender [17]. Sociodemographic details like gender and age can be gathered for members; however, data for non-members will be absent [19]. ...
... Tuli and Mitra collected the map of land use categories from the City of Portland and, using ArcGIS Pro, calculated the land use entropy index to understand how shared e-scooter use is influenced by land use [41]. Yang et al. utilized land use covariates such as residential, commercial, industrial, and park areas, drawing data from the Chicago data portal [27]. ...
... They also considered the exposure to NO 2 (μg/m 3 ), daytime noise (dB), and Normalized Difference Vegetation Index (NDVI) of each trip. Yang et al. incorporated transportation infrastructure covariates such as bike trail density, major road density, the number of docks at nearby bike stations, and the average daily traffic volume (AADT) of the road nearest to each bike station, drawing data from the Chicago data portal [27]. ...
Chapter
In micromobility studies, data plays an important role, enabling the assessment of many aspects of mobility. Various data types are used to explore areas such as safety, policy evaluation, urban planning, and environmental sustainability. This chapter reviews the primary data types, sources, and collection methods in micromobility studies, including sensor data, surveys, field observations, built environment data, and archival data sources. Sensor data, such as mobile phone GPS and vehicle sensors, provide detailed insights into mobility patterns and environmental conditions but lack socio-demographic information. Surveys and observations are the primary data sources for user behavior and use of infrastructure. Built environment data examines factors like density, diversity, and design influencing micromobility. Archival data, including media reports and public records, are crucial for policy analysis and safety evaluations. The chapter also includes common practices in data preprocessing to enhance data accuracy, supporting researchers in advancing micromobility studies.
... rs, and the field of transportation is no exception. New micromobility services have been introduced in recent years, reshaping how people move using evening hours. Notably, in specific urban areas such as Chicago, IL, this peak is observed later in the afternoon, precisely between 16:00 and 20:00 on weekdays, as detailed by Yang's research in 2021(H. Yang et al., 2021 It is important to note that micromobility ridership tends to be higher on days with favorable weather conditions, such as higher temperatures and good visibility, as well as on holidays and days featuring special events. Conversely, e-scooter ridership tends to be lower on days with rainfall, snow, freezing conditions, higher humidity, ...
... Similarly, research conducted in Austin, Texas, by Jiao and Bai (2020) concurred that regions with high population density and lower income levels tend to see increased e-scooter activity. Additionally, areas characterized by higher employment density generated more e-scooter trips and acted as attractive destinations for such journeys ( Caspi et al., 2020 ;Huo et al., 2021). Studies have consistently suggested that e-scooters are not primarily employed in high-income areas but rather find significant utility in low-income and educated communities (Bai and Jiao, 2020;Caspi et al., 2020 ;Merlin et al., 2021 ). ...
... Studies have consistently suggested that e-scooters are not primarily employed in high-income areas but rather find significant utility in low-income and educated communities (Bai and Jiao, 2020;Caspi et al., 2020 ;Merlin et al., 2021 ). Furthermore, these studies align in reporting that micromobility ridership is more prominent in areas with a higher proportion of young males, particularly within the 18-29 age group (Bai and Jiao, 2020;Caspi et al., 2020 ;Hosseinzadeh et al., 2021a ;Huo et al., 2021). Among these studies, Huo et al. (2021) revealed a strong correlation between high e-scooter ridership and a greater percentage of commuters using public transport, as well as a higher proportion of households without cars. ...
... Especially in larger cities, the inner city mobility has been influenced by e-scooter sharing services from various providers commonly founded between 2017 and 2019 [2,3]. Nowadays, devices are used in 630 cities in over 53 countries, making them the most popular form of electronic-powered sharing devices [4,5]. E-scooters are usually quickly available and easy to handle. ...
... 158 patients (86.3%) got injured without wearing a helmet while driving. 25 patients (13.7%) got (6) Plate osteosynthesis (6) Proximal humerus fracture (2) Plate osteosynthesis (1) Intramedullary nailing (1) Shoulder dislocation ± Bankart lesion (2) Arthroscopic surgery (2) AC joint dislocation (2) Joint reconstruction surgery (2) Supraspinatus tendon tear (1) Arthroscopic surgery (1) Hand (10) Finger fracture or dislocation (5) Kirschner wire fixation (3) Screw osteosynthesis (2) Metacarpal fracture (4) Kirschner wire fixation (4) Carpal fracture (1) Screw osteosynthesis (1) Elbow (9) Distal humerus fracture (3) Double plate osteosynthesis (2) Screw osteosynthesis (1) Elbow dislocation (3) Open capsule-ligament repair (3) Olecranon fracture (2) Tension-band wiring technique (2) Radial head fracture (1) Screw osteosynthesis (1) ...
Article
Full-text available
Purpose During the last few years, the number of electric scooter (e-scooter) users has risen to an all-time high. This study aimed to analyze e-scooter related accidents and trauma prevention measures in a large European city (Vienna, Austria). Methods This retrospective study comprises a thorough data assessment and analysis of all e-scooter related accidents between 2018 and 2021 at a large level 1 trauma center in Vienna. Based on the data analysis, risk factors were identified, and possible prevention strategies were proposed. Results During the observed period, 1337 patients sustained an injury from an e-scooter. Of these, 1230 were injured directly while driving (92%). The remaining 107 patients (8%) were classified as non-driving injuries. 927 injuries involved males (69.3%). The mean age was 32.1 years (range 4–86 years). Of all injured patients, 429 (32.1%) sustained at least one serious injury. The most common injuries included radial head fractures and concussions. Among the accidents treated, the use of protective equipment was sporadic. For example, helmets were worn in only 13.7% of cases. Wearing a helmet reduced the number of head injuries (24% versus 46.8%). In just three years, the number of patients increased 19-fold with a focus in the summer months. Conclusion This study shows a substantial and sustained increase in e-scooter accidents with potentially serious injuries. Helmet use was found to be an effective form of head injury prevention. Further options for using protective equipment should be evaluated to improve the safety aspects of riding e-scooters.
... Although the first shared e-scooter system was introduced in 2017 in the US (Yang et al., 2021), it was not introduced in Turkey until late 2019 (Koca, 2019). It has been used by individuals more and more during the COVID pandemic as an alternative to public transport (Erbas, 2020). ...
... Furthermore, an increasing number of studies have begun to focus on the interaction between diferent modes of transportation [32,34]. Qiao explored the competitive and cooperative relationship between ride-sourcing services and public transportation from the perspective of afordability, revealing a complementary relationship between the two [35]. ...
Article
Full-text available
Understanding the factors influencing ride-sourcing trips is crucial for enhancing the quality of personalized mobility and optimizing the allocation of transportation system resources. However, the nonlinear effects of dockless bike-sharing (DBS) and the built environment (BE) across different spatiotemporal contexts have not been adequately addressed in previous research. This study aims to bridge this gap by analyzing order data and BE characteristics in Tianjin, China. Utilizing the Gradient Boosting Decision Tree (GBDT) model and Accumulated Local Effects (ALE) plots, this study explores the relative importance and nonlinear thresholds of these factors on ride-sourcing trips. The findings reveal that DBS trips during weekday AM peak hours exert significantly negative effects on ride-sourcing, whereas the impact during weekend AM peaks and daily PM peaks is positive. Furthermore, variables such as active population density, metro accessibility, and residential, entertainment, and cultural BEs have positive nonlinear impacts on ride-sourcing trips. These insights offer policy implications and resource allocation recommendations for both government bodies and operators.
... This also contradicts the claims expressed in industry and media reports that bike sharing could be losing its market to scooter sharing services (Mobilne Miasto 2020; Wiewiora 2021). Although scooter sharing services have been found to be at least to some extent competitive to BSS (Liu and Lin 2022;Yang et al. 2021), as sources of financing of these two modes of transportation in Poland differ it couldn't be a major reason for Nextbike Polska's problems. Additionally, our results show that Nextbike's financial decline started before the wide popularization of e-scooter sharing in Poland. ...
Article
Full-text available
Urban transportation has changed dramatically in the recent years through the large scale implementation of shared micro-mobility services, especially bike sharing systems (BSS) and electric scooter sharing (ESS). The COVID-19 pandemic brought further changes and uncertainty to this turbulent business environment. In 2020, the major BSS operator in Poland filed for bankruptcy claiming that its problems arise from the COVID-19 pandemic. Market reports and media speculated that BSS business, despite being publicly financed, and considered to be a of part of public transportation system, could be unsustainable in face of the competition from ESS. We used Z‑score analysis to investigate if bike sharing systems operators’ problems began before or during the COVID-19 pandemic and large scale development of ESS. Our study focuses on the Nextbike company, which held a dominant stake in the Polish BSS market. It also covers two other major Polish operators and includes a German operator for comparative analysis. To complement the quantitative findings from z‑score analysis, we have also interviewed representatives of major stakeholders, which deepened our understanding of BSS problems. The results of our research indicate that although the market was affected by the COVID-19 pandemic and growing scooter-sharing competition, the problems of the largest BSS operator were specific to this company, and the entire market was not under the risk of failure.
... Shared micromobility has gained popularity in recent years as a convenient and sustainable mode of transportation that has great potential to enhance transportation accessibility, improve urban air quality, and promote active transportation and physical activity (NABSA, 2022). The potential benefits of shared micromobility have motivated a burgeoning body of literature examining its user profiles and usage patterns (Christoforou et al., 2021;Guo and Zhang, 2021;Hu et al., 2021;Laa and Leth, 2020;Reck and Axhausen, 2021), modal shift dynamics (Bieliński et al., 2021;Fukushige et al., 2021;Kong et al., 2020;Lee et al., 2021;Ma et al., 2019;Martin and Shaheen, 2014;Yan et al., 2021;Yang et al., 2021), and factors influencing its utilization (Bai and Jiao, 2020;Caspi et al., 2020;El-Assi et al., 2017;Hosseinzadeh et al., 2021a;Huo et al., 2021;Jin et al., 2023;Tuli et al., 2021;Yang et al., 2022). The emergence of e-scooter sharing has inspired many comparative studies investigating the different temporal and spatial usage patterns between bikesharing and e-scooter sharing (Hosseinzadeh et al., 2021b;Younes et al., 2020;Zhu et al., 2020). ...
Article
Full-text available
Shared micromobility in the U.S. has rebound after the decline caused by the COVID-19 pandemic, with a substantial increase in the adoption of shared e-bikes nationwide. However, research on hybrid e-bike sharing, which combines station-based and dockless systems, is limited. This study addresses this gap by comparing spatial determinants of hybrid e-bike and dockless e-scooter sharing link flows in 32,965 street segments in Portland, Oregon during 2022, using gradient boosting decision tree (GBDT) models. Distance to the city center emerges as the most important determinant for both modes, with closer proximity to the city center associated with higher link flows. Factors such as the presence and types of bike facilities, the availability of streetlights and street trees, and job density also significantly influence e-bike and e-scooter link flows. A notable difference between the two modes is that e-scooter trips are more sensitive to distance to the city center than e-bike trips. Furthermore, bike facilities have a greater impact on e-bike link flows, whereas job density is more influential in determining e-scooter link flows. These findings offer strategies for policymakers and urban planners to promote and manage shared micromobility and optimize the built environment. These strategies include enforcing higher device availability requirements in underprivileged neighborhoods, transitioning e-scooter sharing systems into a hybrid model, expanding the off-street bike trial network and bikeway network, and augmenting the coverage of streetlights and street trees along the bikeway network.
Article
With the rapid development of shared e-scooters, it is essential to understand their usage patterns for formulating informed e-scooter fleet management policies. This study first analyzes the usage pattern of shared e-scooters in Indianapolis, USA, by mining big e-scooter trip data. The analysis reveals an oversupply of shared e-scooters relative to actual user demand. Thus, a minimum fleet sizing algorithm is proposed to determine the required minimum e-scooter fleet size with the objective of reducing total operation cost, while ensuring demand coverage. Furthermore, three heuristic algorithms are proposed to address the static e-scooter rebalancing problem, focusing on minimizing rebalancing distance cost and rebalancing time. These algorithms consider practical operational constraints, including the number of rebalancing vehicles, their capacity, and the frequency of visits to e-scooter stations by rebalancing vehicles. The proposed algorithms are applied to e-scooter rebalancing scenarios with comparison between the minimum and actual fleet sizes. The case study results in Indianapolis, USA demonstrate that the rebalancing distance cost with the minimum fleet size is significantly lower than that with the actual fleet size. What’s more, the rebalancing time can be reduced by about 12.34% to 27.80% when using the minimum fleet size. The findings of this study offer valuable policy implications and managerial insights for shared e-scooter operators and policymakers in developing effective e-scooter management strategies.
Article
Full-text available
Electric scooter (e-scooter) sharing systems (ESSs) have been widely adopted by many cities around the world and have attracted a growing number of users. Although some studies have explored the usage characteristics and effects of the built environment on ESS ridership using one city as an example, few studies have considered multiple cities to obtain generalizable and robust results. To fill this research gap, we collect the ESS trip data of five cities in the U.S., namely Austin, Minneapolis, Kansas City, Louisville, and Portland, and explore the effects of the built environment on ESS ridership after controlling for socioeconomic factors. The temporal distributions of e-scooter ridership of different cities are similar, having a single peak period on weekdays and weekends between 11:30 and 17:30. In terms of spatial distribution, the ESS ridership is higher in universities and urban centers compared to other areas. Multilevel negative binomial model results show that ESS trips are positively correlated with population density, employment density, intersection density, land use mixed entropy, and bus stop density in the census block group. E-scooter ridership is negatively correlated with the median age of the population in the census block group and distance to the city center. The findings in this article can help operators understand the factors that affect the ridership of shared e-scooters, determine the changes in ridership when the built environment changes, and identify high-ridership areas when ESS is implemented in new cities.
Article
Full-text available
Waiting time (WT) is an important measure that can reflect accessibility to ridesourcing service. Previous studies explored the effects of built environment factors on WT based on estimated WT but did not control for trip-level characteristics, which may lead to biased parameter estimation. Thus, we further study this topic by using the actual WT recorded by the RideAustrin platform and considering trip-level variables. The single-level and multilevel proportional hazards models are constructed, and model comparison shows that the multilevel model performs better. We find that waiting time is positively correlated with trip-level characteristics such as traffic conditions, surge multiplier, and rainy weather. Regarding built environment factors, WT is positively related to distance to CBD and negatively related to road density, transit stop density, and land-use entropy. WT is also higher in areas with a high fraction of Hispanic/Latino and Black residents but lower in areas of low income.
Article
Full-text available
Shared micromobility services (e-scooters, bikes, e-bikes) have rapidly gained popularity in the past few years, yet little is known about their usage. While most previous studies have analysed single modes, only few comparative studies of two modes exist and none so-far have analysed competition or mode choice at a high spatiotemporal resolution for more than two modes. To this end, we develop a generally applicable methodology to model and analyse shared micromobility competition and mode choice using widely accessible vehicle location data. We apply this methodology to estimate the first comprehensive mode choice models between four different micromobility modes using the largest and densest empirical shared micromobility dataset to-date. Our results suggest that mode choice is nested (dockless and docked) and dominated by distance and time of day. Docked modes are preferred for commuting. Hence, docking infrastructure for currently dockless modes could be vital for bolstering micromobility as an attractive alternative to private cars to tackle urban congestion during rush hours. Furthermore, our results reveal a fundamental relationship between fleet density and usage. A "plateau effect" is observed with decreasing marginal utility gains for increasing fleet densities. City authorities and service providers can leverage this quantitative relationship to develop evidence-based micromobility regulation and optimise their fleet deployment, respectively.
Article
The rapid growth of dockless bike-sharing (DBS) systems has attracted increased academic attention in the solutions to first- and last-mile problems. However, only a few studies have examined how the synergy between DBS and metro transit is affected by objective and perceived measures of built environment collectively. This study intends to fill this research gap by focusing on the effects of objective and perceived measures of built environment on DBS–metro integrated use for commuting trips. Results reveal that low agreement between the two measures of built environment and that the perceived measure is more likely to be directly associated with DBS–metro integration than the objective measure. Different built environment attributes may affect DBS–metro integration by unique paths. Moreover, individual characteristics (i.e., gender, age, and income) and location factor moderate the association between the built environment and DBS–metro integration. Particularly, built environment attributes related to transportation service are easier to be moderated than land use and cycling condition attributes. We conclude that the understanding of and interventions for the built environment as objectively measured are necessary but not sufficient for DBS–metro integration. Promoting the perception of the built environment among different population groups is also important for interventions.
Article
Shared micromobility programs, including dockless electric scooter-share (E-scooter), are popular in many U.S. cities, and with their adoption brings the hope that they may uphold better car-free accessibility. However, few studies provide clear answers to what activities drive its travel demand or whether it could actually generate more visiting activities. To fill this gap, we conducted a spatiotemporal similarity analysis between E-scooter use and visit patterns to leisure facilities. We find that E-scooter use is significantly correlated with daily dining and drinking, shopping, and recreational activities, in that order. Moreover, we find higher scooter-visit correlation clusters in downtown and university campus areas. We then used the Difference-in-Differences approach to examine if E-scooter use can generate more visiting activities. Surprisingly, the results show that E-scooter use is insignificant to the overall visit increase.
Article
Shared micro-mobility services have rapidly gained popularity yet challenged city administrations to develop adequate policies while scientific insight is largely missing. From a transportation equity perspective, it is particularly important to understand user correlates, as they are the beneficiaries from public investment and reallocation of public space. This paper provides an up-to-date account of shared micro-mobility adoption and user characteristics in Zurich, Switzerland. Our results suggest that shared micro-mobility users tend to be young, university-educated males with full-time employment living in affluent households without children or cars. Shared e-scooter users, in particular, are younger, yet more representative of the general population in terms of education, full-time employment, income and gender than bike-sharing users. This suggests that shared e-scooters may contribute to transportation equity, yet their promotion should be handled with care as life-cycle emissions exceed those of bike-sharing and equity contributions might be skewed as many users are students.
Article
Micromobility, including the use of shared electric scooters (e-scooters), emerged rapidly in North America and is marketed as an alternative to car reliance, especially for short distance travel in urban settings. Our study aims to contribute to our understanding of how shared e-scooters are used by examining the factors that determine the presence of e-scooters, as well as those that cause variation in e-scooter presence between each consecutive hour and throughout the day. The object of this study is to investigate how temporal, land use, transport infrastructure, and weather attributes impact available e-scooter distribution and variation in e-scooter presence in Washington D.C., to reveal use patterns and develop a framework for studying citywide e-scooter systems. Data on the location of e-scooters in the Washington D.C. area over six full days was collected. Then, multilevel mixed effects linear regression models were generated to investigate the impact of time, land use characteristics, and the built environment while controlling for weather conditions. We found that temporal effects were present, as weekends and late nights were associated with fewer e-scooters and less variation in hourly e-scooter presence. We observed that the average number of e-scooters available per 0.07 mile² on weekends was 0.26 (7.81%) fewer than on weekdays, and 0.82 (24.62%) fewer during the late night than other times of day, all else held constant. Higher population density, density of places of interest, and activities were generally associated with more e-scooters and contributed to more change in the hour-to-hour numbers of e-scooters but less variation throughout the day. Bikeshare stations and bicycle lanes positively impacted presence, they increased the odds of e-scooter presence by 3.16 and 2.73 times respectively and change in the average number of e-scooters nearby. The hourly change in the average numbers of e-scooters near bikeshare stations was 0.19 all else held equal, and it is unclear whether e-scooters were used as first-mile last-mile solutions for public transport. These findings can help policy-makers in cities with comparable climates, land use characteristics, and transport infrastructure. The findings can help city planners and engineers make appropriate decisions in recognizing e-scooters as an urban mobility solution, where to expect them to emerge in different parts of the city, and how e-scooters interact with established transport systems.
Article
Micromobility and especially e-scooter sharing have recently attracted a lot of attention, due to the rapid spreading of e-scooters in many cities around the world. However, many local authorities have not yet been prepared for efficiently integrating e-scooters in their transport systems and the exact impact of e-scooters is still unclear. It is therefore essential to understand the way e-scooters operate and their users’ profile. To address these questions, a study was designed based on 578 questionnaires (271 by e-scooter users and 307 by non-users) in the city of Thessaloniki, Greece. The analysis utilized a classification tree model for identifying the characteristics of people that are attracted by e-scooters (i.e., used them more than once) and a latent variable logit model for understanding the attributes of the regular e-scooter users. The results show that shared e-scooters mostly replaced walking and public transport trips; therefore, the positive impact of e-scooters on the environment is questioned. Also, the results indicate that people traveling with bicycle or motorcycle were not at all attracted by e-scooters. Moreover, females seem to be less keen on using e-scooters compared to males, while people living downtown are more regular users compared with those living in longer distances from the city center. These findings can aid policymakers in shaping the manner with which e-scooters can be incorporated in their cities.
Article
In recent years, many cities have implemented bike-sharing programs (BSP) to improve the travel efficiency of short trips. Early studies have analyzed how built environment factors affected bike-sharing usage. However, these studies mainly used global regression models that cannot demonstrate the spatial variation relationship between the built environment and bike usage. Therefore, this study employs both a global regression model and a geographically weighted regression (GWR) model to examine the global and local influences of the built environment on bike usage, which represents the average bike trips on workdays and non-workdays. This research takes Suzhou, China as a case study area. It uses one-year bike-sharing trip data, metro ridership data, cycling infrastructure data, cellular signaling data, and points of interest (POI) data. The global regression results show that bike stations near public transit, restaurants, shopping malls, and educational and financial places have high numbers of bike trips on both workdays and non-workdays; however, bike station proximity to workplaces is positively associated with bike trips on workdays but not on non-workdays. The results of GWR are partially consistent with the global regression results and show the local effects of the built environment on bike usage in different parts of Suzhou. Also, the goodness of fit in the GWR is better than that of the global regression model. The findings of this study provide strategic guidance to improve the service quality of bike-sharing systems as there is a pressing need to integrate BSP policies into the land use planning framework to encourage more diverse transport modal change and incentivize more commuting.
Article
Shared e-scooter systems are operating across hundreds of cities worldwide. However, limited understanding of the user demand, as well as how this demand varies across individuals with various transportation preferences living in different urban contexts, is a key barrier to developing policy and regulations. This paper begins to close this gap by providing preliminary but novel insights into the socio-demographic, attitudinal and built-environment characteristics of potential users of shared e-scooters. In particular, we examine the self-reported intention to consider shared e-scooters by residents in Toronto and surrounding municipalities in Canada. Based on an online survey of 1,640 adults living in 17 neighbourhoods, we found that 21% were amenable to considering e-scooters for some of their current trips, and the majority would replace their existing walking (60%) and transit (55%) trips with shared e-scooters. Weighted logistic regression models revealed that all else being equal, preference toward trip efficiency, and environment and health-consciousness, were positively associated with potential e-scooter consideration. Perceived walkability/bikability and street safety also increased the likelihood of considering shared e-scooter in future. The findings begin to identify who will likely benefit from this micro-mobility option and where the impacts will be felt the most.