Content uploaded by Daniel Jian Sun
Author content
All content in this area was uploaded by Daniel Jian Sun on Oct 02, 2019
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=gcpi20
Traffic Injury Prevention
ISSN: 1538-9588 (Print) 1538-957X (Online) Journal homepage: http://www.tandfonline.com/loi/gcpi20
Taxi driver speeding: Who, when, where and how?
A comparative study between Shanghai and New
York City
Yizhe Huang, Daniel (Jian) Sun & Juanyu Tang
To cite this article: Yizhe Huang, Daniel (Jian) Sun & Juanyu Tang (2018) Taxi driver speeding:
Who, when, where and how? A comparative study between Shanghai and New York City, Traffic
Injury Prevention, 19:3, 311-316, DOI: 10.1080/15389588.2017.1391382
To link to this article: https://doi.org/10.1080/15389588.2017.1391382
View supplementary material
Accepted author version posted online: 18
Oct 2017.
Published online: 09 Feb 2018.
Submit your article to this journal
Article views: 136
View related articles
View Crossmark data
TRAFFIC INJURY PREVENTION
, VOL. , NO. , –
https://doi.org/./..
Taxi driver speeding: Who, when, where and how? A comparative study between
Shanghai and New York City
Yizhe Huanga,b, Daniel (Jian) Suna,b,c, and Juanyu Tangb,c
aChina Institute of Urban Governance, Shanghai Jiao Tong University, Shanghai, China; bCenter for ITS and UAV Applications Research, Shanghai Jiao
Tong University, Shanghai, China; cSchool of Naval Architecture Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai, China
ARTICLE HISTORY
Received June
Accepted October
KEYWORDS
Taxi driver; speeding; DREI
method; GPS data;
comparative analysis
ABSTRACT
Objective: The 3 objectives of this study are to (1) identify the driving style characteristics of taxi drivers in
Shanghai and New York City (NYC) using taxi Global Positioning System (GPS) data and make a comparative
analysis; (2) explore the inuence of dierent driving style characteristics on the frequency of speeding (who
and how?) and (3) explore the inuence of driving style characteristics, road attributes, and environmental
factors on the speeding rate (when, where, and how?)
Methods: This study proposes a driver–road–environment identication (DREI) method to investigate the
determinant factors of taxi speeding violations. Driving style characteristics, together with road and envi-
ronment variables, were obtained based on the GPS data and auxiliary spatiotemporal data in Shanghai and
NYC.
Results: The daily working hours of taxi drivers in Shanghai (18.6 h) was far more than in NYC (8.5 h). Theaver-
age occupancy speed of taxi drivers in Shanghai (21.3 km/h) was similar to that of NYC (20.3 km/h). Speeders
in both cities had shorter working hours and longer daily driving distance than other taxi drivers, though
their daily income was similar. Speeding drivers routinely took long-distance trips (>10 km) and preferred
relatively faster routes. Length of segments (1.0–1.5 km) and good trac condition were associated with
high speeding rates, whereas central business district area and secondary road were associated with low
speeding rates. Moreover, many speeding violations were identied between 4:00 a.m. and 7:00 a.m. in both
Shanghai and NYC and the worst period was between 5:00 a.m. and 6:00 a.m. in both cities.
Conclusions: Characteristics of drivers, road attributes, and environment variables should be considered
together when studying driver speeding behavior. Findings of this study may assist in stipulating relevant
laws and regulations such as stricter oense monitoring in the early morning, long segment supervision,
shift rule regulation, and working hour restriction to mitigate the risk of potential crashes.
Introduction
Speeding is one of the most signicant contributors to traf-
c accidents in many countries, such as China (Trac Man-
agement Bureau 2012) and the United States (National Cen-
ter for Statistics and Analysis 2013;Royal2003; Schroeder et
al. 2013). Many literatures have studied the factors aecting
driver’s speeding behavior on both highways and urban arte-
rials (Harre 1996;Giles2004;Lefeve1956; Oppenlander 1966;
Rakauskas et al. 2007;Richardetal.2013;Williamsetal.2006;
Zhang et al. 2014). However, limited studies have been con-
ducted on the issue of taxi speeding. Compared to ordinary
drivers, taxi drivers are more likely to commit risk behaviors
(Burns and Wilde 1995;Mayhew2000;Tseng2013;Yehetal.
2015). Between 1996 and 2000 in New South Wales, Australia,
7,923 taxi drivers were involved in crashes, nearly 10% (n=
750) were killed or injured (Lam 2004). In South Africa, a study
showed that 33.8% of taxi drivers had been involved in a car acci-
dent (Peltzer and Renner 2003), whereas another study in Hanoi
reported that 276 of 1,214 taxi drivers (22.7%) were involved
CONTACT Daniel (Jian) Sun danielsun@sjtu.edu.cn China Institute of Urban Governance and Center for ITS and UAV Applications Research, Shanghai Jiao Tong
University, A, Ruth Mulan Chu Chao Building, No. Dongchuan Road, Shanghai , China.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/gcpi.
Associate Editor Douglas J. Gabauer oversaw the review of this article.
Supplemental data for this article can be accessed on the publisher’s website.
in at least one crash (La et al. 2013). Studying taxi speeding is
important because the time on the road of taxi drivers is consid-
erable (Dalziel and Job 1997) and they have more opportunities
to be involved in a speeding violation (Tseng 2013).
Only a few studies have focused on the taxi drivers’ speed-
ingproblem.Inanobservationalstudy,BurnsandWilde(1995)
foundthattaxidriverswithasensationseekingpersonalitywere
more frequently convicted of speeding violations. Newnam et
al. (2014) utilized questionnaires and found that age and edu-
cational level had no signicant relationship with taxi drivers’
speeding behavior in Ethiopia. Using the data from a national
survey in Taiwan, Tseng (2013) explored the determinant per-
sonal factors of taxi drivers’ speeding violations and found that
age, job experience, operation styles, daily driving distance, driv-
ing late at night, and monthly o-duty days were signicantly
associated with speeding violations. According to the same sur-
veysourceinTaiwan,Yehetal.(2015)analyzedthefactorsof
female taxi drivers’ speeding oenses and found that female
taxi drivers’ speeding oenses were signicantly related to age,
© Taylor & FrancisGroup, LLC
312 Y. HU ANG E T A L.
education level, and mileage driven, whereas job experience,
business operating style, and vehicle engine size did not sig-
nicantly aect the percentage who had at least one speeding
ticket. Shi et al. (2014) designed questionnaires to explore the
factors aecting aberrant behavior of taxi drivers in Beijing.
They found that economic pressure, ownership of taxi, and the
complaint system would aect taxi drivers’ aberrant behaviors,
including speeding. However, these taxi driver speeding studies
neglected some important factors. For ordinary drivers, previ-
ous studies have demonstrated that speeding behaviors are also
triggered by exogenous pressure, such as road attributes (Giles
2004;Zhangetal.2014), trac conditions (Zhang et al. 2014),
vehicle parameters (Giles 2004;Williamsetal.2006;Zhanget
al. 2014), time of day (Oppenlander 1966;Richardetal.2013;
Zhang et al. 2014), light conditions (Lefeve 1956;Zhangetal.
2014), and location factors (Giles 2004; Rakauskas et al. 2007).
However, whether these situational factors lead to speeding vio-
lations remain unclear for taxi drivers.
In summary, with respect to taxi speeding, driving style char-
acteristics and situational factors have not been adequately stud-
ied, largely due to the scarcity of reliable taxi driver speeding-
related data. Data required to explore the factors associated with
taxi speeding include demographic characteristics, road attri-
butions, vehicle parameters, and other environmental variables.
Most previous studies (La et al. 2013;Shietal.2014;Tseng2013;
Ye h e t a l . 2015) have used survey data to investigate driver-
related variables. The conventional methods were easy to man-
age and were useful to identify the demographic characteristics.
However, the results may have self-reported biases and are insuf-
cient to explore situational factors. The present study explored
these determinant factors that lead to taxi speeding using
Floating Car Data (FCD) methods. Some studies (Sun et al.
2014;Nianetal.2017;Zhangetal.2017) have demonstrated the
advantages of implementing FCD methods in taxi trac vio-
lation research. Compared to survey studies, the FCD method
is capable of capturing spatial–temporal information on taxi
speeding behavior. The sample size of more than 10,000 Global
Positioning System (GPS)-equipped taxis can improve the accu-
racyofthestudy.The3objectivesofthisstudyareasfollows:
rTo identify the driving style characteristics of taxi drivers
in Shanghai and New York City (NYC) using taxi GPS data
and make a comparative analysis.
rTo explore the inuence of dierent driving style charac-
teristics on the frequency of speeding (who and how?).
rTo explore the inuence of driving style characteristics,
road attributes, and environmental factors on the speeding
rate (when, where, and how?).
Methodology
Overview of the DREI method
This section provides the driver–road–environment identica-
tion (DREI) method to explore the relationships among driving
style characteristics, roadway attributes, and environment fac-
tors associated with speeding for taxi drivers. The method con-
sists of 3 steps: (a) ltering the data and matching the data on the
map; (b) identifying driving style characteristics and compar-
ing the basic characteristics between speeders and ordinary taxi
drivers in 2 cities; (c) adding the situational data and evaluating
the signicance and strength of driving style characteristics and
other situational variables on speeding rate.
Data
The GPS data for Shanghai were collected in 2015, with 13,475
taxis for one month (April 1–30, 2015). The data on April 10
(Friday) were used to identify the characteristics of taxi driving
style and speeding behavior. The GPS data for NYC were col-
lected in 2013. The data on April 12 (Friday) were adopted to
identify speeding behavior.
Measure and variables
Comparisons of taxi speeding in Shanghai and NYC: Who is
speeding? And who speeds the most?
The speed choice of a driver was fairly consistent over time
(Haglund and ˚
Aberg 2000;Ahieetal.2015)andwashighly
inuenced by their usual speeds (Ahlin 1979). In this article, we
dened taxi drivers who frequently travel faster than other taxi
drivers as speeders. These drivers would be more likely to get a
better position or speed advantage whenever they have a chance.
As a control group, ordinary taxi drivers were randomly selected
fromtheentiretaxidriversample.
Previous studies demonstrated that daily working hours,
daily driving distance, driving late at night, and weekly days
o (Tseng 2013;Walietal.2017) would signicantly aecting
taxi drivers’ speeding behaviors. However, the business operat-
ing style of taxi drivers was a controversial factor aecting taxi
speeding (Tseng 2013;Yehetal.2015). In this study, the busi-
ness operating style was explored using long-distance trip ratio,
route preference, time occupancy rate factors, and daily prot
of drivers. Many taxi drivers in Shanghai prefer long-distance
trips, such as from/to an airport, because they can earn the long-
distance surcharge when the trip is longer than 10 km (no such
surcharge in NYC). It is possible that these drivers may have
a higher probability to commit a speeding violation in a long-
distance trip. In addition, it is doubtful that drivers with dier-
ent route preferences would have dierent speeding behaviors.
A previous study (Liu et al. 2010)showedthatsometaxidrivers
may choose the shortest route from the origin to the destina-
tion under occupancy status, whereas others tend to choose the
relatively faster path to deliver passengers. In order to exam-
ine the eect of route preference, the route directness index is
chosen to reect the circuity of the route; the calculation for-
mula to calculate the route directness index is from Cui et al.
(2016). Considering that taxi drivers with dierent operational
performances have dierent working experience and working
pressure,itispossiblethatdriverswithvariousoperationalper-
formances would commit dierent speeding violations. There-
fore, the operation performance of taxi drivers was investigated
through daily prot and time occupancy rate. The time occu-
pancy rate is calculated as the passenger occupancy time divided
by the total working time.
The present study is to explore the dierence of these driv-
ing style characteristics between speeders and ordinary taxi
drivers in Shanghai and NYC to answer the question, “Who
is speeding in Shanghai and NYC?” Then, for the question of
who speeds the most, a stepwise regression was employed to
evaluate the relative signicance and strength of the inuential
TRAFFIC INJURY PREVENTION 313
Tab le . Basic characteristics of ordinary taxi drivers and speeders in Shanghai and NYC.
Shanghai NYC
Characteristics Ordinary taxi driver (N=,) Speeder (N=,) Ordinary taxi driver (N=,) Speeder (N=,)
Daily working hours (h) . . . .
Driving late at night .% .% .% .%
Income (US dollars) . . . .
Daily driving distance (km) . . .a.a
Daily occupancy driving distance (km) . . . .
Weekly time off (days) .b.b. .
Weekly working hours (h) . . . .
Average occupancy speed (km/h) . . . .
Time occupancy rate . . . .
Long-distance rate . . . .
Route directness index . . . .
aEstimated by setting the same rate of average occupancy speed and idle speed as Shanghai (.).
bSet to ., because each Shanghai taxi generally has alternative drivers.
characteristics variables on speeding violations to eliminate
the eect of multicollinearity among variables. Likewise, the
results of Shanghai and NYC were compared to determine the
common and dierent characteristic variables inuencing taxi
driver speeding oenses.
Road-level analysis in Shanghai: When, where, and how?
Previous studies explored the determinant factors leading to
speeding behavior and demonstrated that time of day (Oppen-
lander 1966;Zhangetal.2014), trac condition (Zhang et al.
2014), location (Rakauskas et al. 2007), and road type and road
length (Giles 2004;Zhangetal.2014) would signicantly inu-
ence drivers’ speeding behaviors. The assumption of the present
study is that these variables are also associated with taxi drivers’
speeding behaviors. Thus, these situational explanatory vari-
ables were included in this study. Considering that the NYC data
set did not include the detailed taxi trajectories on the road,
we performed this analysis using data only from Shanghai. The
areacoloredinyellowinFigureA1(seeonlinesupplement)
waschosenasourstudyarea,whichcontainedtheouterringof
expressways and 2 main airports in Shanghai. In total, 265 arte-
rial roads and 421 secondary roads segments longer than 300 m
were included.
By matching the GPS traces on the road, spatiotemporal situ-
ational features can be identied, such as time of day, road type,
road length, speed limits of the road segments, and regional
information. The trac condition of each road segment can be
roughly estimated by the average speed and speed variation of
other taxis driving on that road segment in a period of time. In
this study, the hourly average speed of each road segment was
considered as an independent sample, which (686 ∗24) were
clusteredinto5groupsbytheiraveragespeedandstandarddevi-
ation features using K-means clustering, as shown in Figure A2
(see online supplement). The proportion of road segments in
each cluster is presented in Figure A3 (see online supplement),
which changes for each single hour.
A generalized linear model (GLM) was employed to iden-
tify how the characteristics and situational factors lead to the
speeding rate. The GLM is a systematic extension of a linear
model for nonnormal data. In this study, the dependent vari-
able yof GLM represented the speeding rate. The explanatory
variables were based on previous studies and the assumption of
this study included daily working hours, daily driving distance,
daily prot, long-distance trip rate, driving late at night, route
directness index, time of day, road speed category, time occu-
pancy rate, road length, location index, and road type.
Result analysis
Comparisons of taxi speeding in Shanghai and NYC: Who is
speeding?
Tabl e 1 presents the basic characteristics of the chosen drivers,
which were calculated by GPS data in Shanghai and NYC. Most
taxi drivers in Shanghai choose to work a 24-h shift and then
rest for an entire day, whereas most taxi drivers in NYC choose
tooperate2shiftsperdayandseldomhaveawholedaytorest.
Although the weekly working hours between Shanghai and NYC
were similar, the daily working hours in Shanghai (18.6 h) were
far more than those in NYC (8.5 h) because of dierent shift
rules. The average occupancy speed of taxi drivers in Shanghai
(21.3 km/h) was similar to that of NYC (20.3 km/h). In both
cities, compared to ordinary taxi drivers, speeders had shorter
working hours but higher driving distances. In addition, the
income of both ordinary taxi drivers and speeders was similar.
Moreover, in both cities, speeders routinely took long-distance
trips and chose relatively fast routes compared to shorter ones.
Figures 1 and 2present the distribution of speeds relative
to the posted speed (PS) for each hour in Shanghai and NYC,
respectively. The category of each column was based on the
speeding penalty classication in Shanghai and NYC (e.g., PS
+20% represented driving at speeds between posted speed and
20% above the posted speed because drivers at that interval
would largely be issued with a speeding ticket); detailed speed-
ing penalty rules are shown in Table A1 (see online supplement).
Figure . Percentages of the different speeding rates for each hour in Shanghai.
314 Y. HU ANG E T A L.
Figure . Percentages of the different speeding rates for each hour in NYC.
All drives that were 20% or more below the posted speed in
Shanghai and 5 mph or more below the posted speed in NYC
were removed. These driving records may not be at the free-ow
time, dened as periods of time that drivers have opportunities
to exceed the speed limit (Richard et al. 2013). It was found that
the speeding feature was time-varying; that is, drivers were more
likely to commit speeding violations between 4 and 7 and com-
mit most between 5 and 6 in both Shanghai and NYC.
Analysis of driver characteristics on speeding frequency:
Who speeds the most?
Stepwise regression was used to explore the signicance
and strength of the driving style characteristics leading to
speeding violations. The results for Shanghai and NYC are
shown in Tables A2 and A3 (see online supplement). Among
the explanatory variables, daily dr iving distance, daily prot, and
long-distance trip rate are relatively important factors signi-
cantly inuencing taxi speeding.
In Shanghai, drivers with long daily working hours and
long driving distances generally had more speeding violations
because of the long duration of exposure on the road. However,
drivers with high daily working hours (>10 h) were less likely to
have speeding violations in NYC (B=−0.05, P=.003). Drivers
whoalwaysdrovelateatnightweremorelikelytospeedinboth
Shanghai (B=16.431, P=.001) and NYC (B=0.041, P=.001).
Moreover, drivers who took long-distance trips in both Shang-
hai (long distance rate >0.4) and NYC (long distance rate >0.5)
were more likely to speed (P<.05). The results also showed that
the variable time occupancy rate signicantly inuenced speed-
ing violations in NYC (P=.001) but not in Shanghai.
GLM analysis of determinant factors: When, where, and
how?
In this section, GLM analysis was employed to further explore
the determinant factors of taxi drivers’ committing speeding
violations. The dependent variable yrepresented the speeding
rate, that is the percentage of speeding over the speed limit. The
explanatory variables included the driver style characteristics
factors and situational factors. The results of GLM analysis are
provided in Tab le 2. The condition number (kappa =12.09) is
less than 100, which indicates that the multicollinearity eect
barely exists in the analysis.
It is indicated that drivers with long working hours (e.g.,
longer than 22 h) were less likely to excess the speed limits at a
high speeding rate (B=−0.056, P=.000), and those with short
working hours (e.g., less than 16 h) were more likely to drive at
higher speeds over the speed limit. The results also showed that
drivers with long daily driving distances (longer than 400 km)
were associated with high speeding rates. Although drivers with
high incomes exceeded the speed limit more often, the results
of GLM analysis indicated that these drivers did not drive at
high speeds over the speed limit. Drivers who routinely took
long-distance trip rate (Long-distance trip rate (LDR) >0.6)
were more likely to drive at high speeding rates (B=0.009, P
=.000). The results of GLM also indicated that driving late at
night and high route directness index were associated with high
speeding rates.
Thespeedingrateoftaxidriverswaslowwhentheroadswere
crowded, with low average speed and variation. The speeding
rate was high when the trac was light, and the highest speeding
rate occurred on roads with the highest speed variation (group 5,
B=0.503, P=.000) rather than the roads with highest average
speed (group 4, B=0.254, P=.000). The temporal data indi-
cated that drivers generally had a higher speeding rate at night
(B=0.028, P=.000). The result of time occupancy rate, another
time-varying indicator, indicated that drivers were more likely
to commit speeding violations during the period with high taxi
demand (time occupancy rate >0.6, B=0.009, P=.000).
Road segments with lengths from 1.0 to 1.5 km contributed
to the highest speeding rates (B=0.042, P=.000). As to
location, the highest speeding rate was more likely to occur
in urban areas (B=0.039, P=.000) and suburban areas
(B=0.006, P=.000) compared to the central business district
(B=0.000). In addition, drivers were less likely to drive at a
high speeding rate on secondary roads compared to arterial
roads (B=−0.291, P=.000).
Discussion
Speeding behavior and determinant factors
Taxi drivers in both Shanghai and NYC had high workloads
and little rest time. The weekly working hours of ordinary taxi
drivers in Shanghai and NYC were 65.1 and 51.9 h, respectively.
This nding was consistent with previous studies that the time
on the road of taxi drivers was often considerable (Dalziel and
Job 1997;Tseng2013;Yehetal.2015). Daily working hours
was another determinant factor for taxi speeding violations in
Shanghai, and drivers with higher working hours were more
likely to commit speeding violations. However, this nding was
not in line with taxi speeding violations in NYC. The reason
may be that taxi drivers in Shanghai needed to work much
harder than NYC taxi drivers during one shift. The working
hours of most taxi drivers in Shanghai were longer than 18 h,
and over 70% Shanghai taxi drivers worked driving late at night
(12:00 a.m. to 6:00 a.m.), whereas the average working hours
of NYC taxi drivers was 8.5 h per day, and less than 40% of
them drove late at night or drove early in the morning. Thus,
under extremely high workload, taxi drivers in Shanghai were
more likely to exceed the speed limit. In both Shanghai and
NYC, the income of speeders and ordinary taxi drivers was sim-
ilar, though speeders tended to have less working hours and
more break time than ordinary taxi drivers. One way to improve
earnings is by increasing working hours, which is consistent
TRAFFIC INJURY PREVENTION 315
Tab le . GLM analysis for estimating speeding rates in Shanghai.
Variables BSE tSig. Variables BSE tSig.
Intercept . . . . Route directness index
Daily working hours (h) <. a———
< a— — — .–. . . . .
– −. . −. . .–. . . . .
– −. . −. . >. . . . .
– −. . −. . Road speed category
> −. . −. . Group a———
Daily driving distance (km) Group . . . .
< a— — — Group . . . .
– . . . . Group . . . .
– . . . . Group . . . .
– . . . . Time occupancy rate
> . . . . <. a———
Daily profit (RMB) .–. . . . .
< a——— >. . . . .
– −. . −. . Road length (km)
–, −. . −. . <. a———
>, −. . −. . .–. −. . −. .
Long-distance trip rate .–. . . . .
<. a— — — .–. . . . .
.–. . . . . >. . . . .
.–. . . . . Location index
.–. . . . . Central business district a———
>. . . . . Urban area . . . .
Driving late at night Suburban area . . . .
No a———Roadtype
Yes . . . . Arterial road a———
Time of day Secondary road −. . −. .
Daytime a— — — Akaike information criterion −,
Night . . . . Kappa .
Dependent variable y=speeding rate.
aSet to .
with ndings in Beijing (Shi et al. 2014) and Hanoi (La et al.
2013).
Many previous studies pointed out that situational factors
are important predictors of ordinary drivers’ speeding viola-
tions (Giles 2004; Oppenlander 1966; Rakauskas et al. 2007;
Williams et al. 2006;Zhangetal.2014). The present study also
indicates that situational factors are determining factors for taxi
drivers’ speeding violations. The speeding rate at night (6:00
p.m.–6:00 a.m.) was signicantly higher than that in the daytime
(6:00 a.m.–6:00 p.m.), and taxi drivers in both Shanghai and
NYC were more likely to commit speeding violations between
4:00 a.m. and 7:00 a.m. and the worst time period was dur-
ing 5:00 a.m. to 6:00 a.m. This nding was in line with pre-
vious studies (Tseng 2013;Zhangetal.2014)thatindicated
that speeding violations were more likely to take place at night.
One possible reason is that trac conditions at night are much
better than in the daytime. The second reason is that driv-
ing at night is less likely to be caught by the police (Tseng
2013).
Itwasalsofoundthatthespeedingrateonarterialroadswas
signicantly higher than on secondary roads, which is consis-
tent with extant research noting that arterial roads in China
were the least safe roads (Zhang et al. 2017). The length of the
road segment was also an important factor aecting taxi drivers’
speeding rates. Taxi drivers were more likely to commit speed-
ing violations at a higher speeding rate on road segments from
1.0 to 1.5 km in length. It is possible that without frequently
being disrupted by crossings or trac signal, drivers were able
to accelerate to a high travel speed in long road segments. This
is consistent with Greibe (2003), who noted that long segments
were associated with high crash frequency. In addition, a high
speeding rate is more likely to take place outside the inner
ring expressways in Shanghai. As indicated in Rakauskas et al.
(2007), drivers in such areas may have lower perceptions of
therisksandperceivealowervalueingovernment-sponsored
trac safety interventions than in central areas.
Limitations
This study has several limitations. First, some important demo-
graphic variables, such as gender, age, and education level, as
well as some important situational variables, such as rain and
seasonal variation, were not included in the previous study. Sec-
ond, the FCD data used in this study were recorded every 10–
30 s, and the time and space resolution of the data may directly
inuence the results. Third, possible interaction eects were
neglectedinthisstudybecauseofthelimitationsofthelinear
regression methods.
Funding
This research was supported in part by the Major Project of National Social
Science Foundation of China (16ZDA048), the Shanghai Municipal Natural
Science Foundation (17ZR1445500), China, and the Humanities and Social
Science Research Project, Ministry of Education (15YJCZH148), China.
Any opinions, ndings and conclusions or recommendations expressed in
this paper are those of the authors and do not necessarily reect the views
of the sponsors.
316 Y. HU ANG E T A L.
References
Ahie LM, Charlton SG, Starkey NJ. The role of preference in speed choice.
Transp Res Part F Trac Psychol Behav.2015;30:66–73.
Ahlin FJ. An Investigation into the Consistency of Drivers’ Speed Choice [M.A.
Sc. thesis]. Department of Civil Engineering, University of Toronto,
Tor o nt o, Can a da; 1979.
Burns PC, Wilde GJ. Risk taking in male taxi drivers: relationships among
personality, obser vational data and driver records. Pers Individ Dif.
1995;18:267–278.
Cui J, Liu F, Hu J, Janssens D, Wets G, Cools M. Identifying mismatch
between urban travel demand and transport network services using
GPS data: a case study in the fast growing Chinese city of Harbin. Neu-
rocomputing.2016;181:4–18.
Dalziel JR, Job RFS. Motor vehicle accidents, fatigue and optimism bias in
taxi drivers. Accid Anal Prev.1997;29:489–494.
Giles MJ. Driver speed compliance in Western Australia: a multivariate
analysis. Transp Polic y.2004;11:227–235.
Greibe P. Accident prediction models for urban roads. Accid Anal Prev.
2003;35:273–285.
Haglund M, ˚
Aberg L. Speed choice in relation to speed limit and inuences
from other drivers. Transp Res Part F Trac Psychol Behav.2000;3:
39–51.
Harré N, Field J, Kirkwood B. Gender dierences and areas of common
concern in the driving behaviors and attitudes of adolescents. JSafety
Res.1996;27(3):163–173.
La QN, Lee AH, Meuleners LB, Duong DV. Prevalence and factors asso-
ciated with road trac crash among taxi drivers in Hanoi, Vietnam.
Accid Anal Prev.2013;50:451–455.
Lam LT. Environmental factors associated with crash-related mortality and
injury among taxi drivers in New South Wales, Australia. Accid Anal
Prev.2004;36:905–908.
Lefeve BA. Relation of accidents to speed habits and other driver character-
istics. Highway Research Board Bulletin.1956;120:6–30.
Liu L, Andris C, Ratti C. Uncovering cab drivers’ behavior patterns from
their digital traces. Comput Environ Urban Syst.2010;34:541–548.
Mayhew C. Violent Assaults on Taxi Drivers: Incidence Patterns and Risk
Factors. Canberra, Australia: Australian Institute of Criminology; 2000.
Report No. 178.
National Center for Statistics and Analysis. Speeding: Trac Safety Facts
2011 Data.Washington,DC:NHTSA,U.S.DepartmentofTransporta-
tion; 2013. DOT HS 811 751.
Newnam S, Mamo WG, Tulu GS. Exploring dierences in driving behav-
ior across age and years of education of taxi drivers in Addis Ababa,
Ethiopia. Saf Sci.2014;68(10):1–5.
Nian G, Li Z, Zhu W, Sun DJ. Analyzing behavior dierences of occupied
and non-occupied taxi drivers using oating car data. J Shanghai Jiao
Tong Univ (Sci).2017;22(6):682–687.
Oppenlander JC. Variables Inuencing Spot Speed Characteristics.Wash-
ington, DC: National Academy of Sciences; 1966.HighwayResearch
Board Special Report No. 89.
Peltzer K, Renner W. Superstition, risk-taking and risk perception of acci-
dents among South African taxi drivers. Accid Anal Prev.2003;35:619–
623.
Rakauskas M, Ward N, Gerberich S, Alexander B. Rural and Urban Safety
Cultures: Human-Centered Interventions toward Zero Deaths in Rural
Minnesota. Minneapolis, MN: Minnesota Department of Transporta-
tion; 2007. Mn/DOT 2007-41.
Richard CM, Campbell JL, Lichty MG, Brown JL, Chrysler S, Lee
JD. Motivations for Speeding, Volume II: Findings Report.Wash-
ington, DC: NHTSA, U.S. Department of Transportation; 2013.
DOT HS 811 818.
Royal D. National Survey of Speeding and Unsafe Driving Attitudes
and Behaviors: 2002, volume II: Findings Report.Washington,DC:
NHTSA, U.S. Department of Transportation; 2003. DOT HS 809
688.
Schroeder P, Kostyniuk L, Mack M. 2011 National Survey of Speeding Atti-
tudes and Behaviors. Washington, DC: NHTSA, U.S. Department of
Transportation; 2013. DOT HS 811 865.
Shi J, Tao L, Li X, Xiao Y, Atchley P. A survey of taxi drivers’ aber-
rant driving behavior in Beijing. JTransSafety&Security.2014;6:
34–43.
Sun DJ, Zhang C, Zhang L, Chen F, Peng Z-R. Urban travel behavior analy-
sis and travel time predicting based on oating car data. Tran sp Lett er s:
In J Trans Res.2014;6(3):118–125.
Trac Management Bureau. Annual Statistical Report on Road Trac
Accidents. Beijing, China: Ministry of Public Security; 2012.
Tseng CM. Operating styles, working time and daily driving distance in
relation to a taxi driver’s speeding oenses in Taiwan. Accid Anal Prev.
2013;52:1–8.
Wali B, Ahmed A, Iqbal S, Hussain A. Eectiveness of enforce-
ment levels of speed limit and drink driving laws and associated
factors—Exploratory empirical analysis using a bivariate ordered
probit model. J Trac Trans Eng (English Edition).2017;4(3):
272–279.
Williams AF, Kyrychenko SY, Retting RA. Characteristics of speeders.
JSafetyRes.2006;37:227–232.
Yeh MS, Tseng CM, Liu HH, Tseng LS. The factors of female taxi drivers’
speeding oenses in Taiwan. Transp Res Part F Trac Psychol Behav.
2015;32:35–45.
Zhang K, Sun DJ, Shen S, Zhu Y. Analyzing spatiotemporal congestion
pattern on urban roads based on taxi GPS data. JTransLandUse.
2017;10(1):675–694.
Zhang G, Yau KKW, Gong X. Trac violations in Guangdong Province
of China: speeding and drunk driving. Accid Anal Prev.2014;64:
30–40.
