Accepted Manuscript

Title: Traffic Violations in Guangdong Province of China:

Speeding and Drunk Driving

Author: Guangnan Zhang Kelvin K.W. Yau Xiangpu Gong

PII: S0001-4575(13)00447-8
DOI: http://dx.doi.org/doi:10.1016/j.aap.2013.11.002
Reference: AAP 3343

To appear in: Accident Analysis and Prevention

Received date: 29-11-2012

Revised date: 16-8-2013
Accepted date: 6-11-2013

Please cite this article as: Zhang, G., Yau, K.K.W., Gong, X.,Traffic Violations in
Guangdong Province of China: Speeding and Drunk Driving, Accident Analysis and
Prevention (2013), http://dx.doi.org/10.1016/j.aap.2013.11.002

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
*Highlights (for review)

Highlights

 We analyze the traffic violations data in China.

 We focus on two specific traffic violation events: speeding and drunk driving.
 Driver, vehicle, road and environmental risk factors are considered.
 Several factors are associated with a significantly higher probability of both speeding and drunk driving:
male drivers, private vehicles, the lack of street lighting at night and poor visibility.
 Other unique risk factors for either speeding or drunk driving are also identified.

t
ip
cr
us
an
M
d
p te
ce
Ac

Page 1 of 24
*Manuscript

Traffic Violations in Guangdong Province of China: Speeding and Drunk

Driving
Guangnan Zhanga, Kelvin K.W. Yaub,, Xiangpu Gonga
a
Center for Studies of Hong Kong, Macao and Pearl River Delta, Sun Yat-Sen University, Xingang Xi
Road, Guangzhou, China
b
Department of Management Sciences, City University of Hong Kong, Tat Chee Avenue, Kowloon,
Hong Kong

t
ip
Abstract

The number of speeding- and drunk driving-related injuries in China surged in the years immediately

cr
preceding 2004 and then began to decline. However, the percent decrease in the number of speeding
and drunk driving incidents (decrease by 22%) is not proportional to the corresponding percent

us
decrease in number of automobile accident-related injuries (decrease by 47%) from the year 2004 to
2010 (Traffic Management Bureau, Ministry of Public Security, Annual Statistical Reports on Road
Traffic Accidents). Earlier studies have established traffic violations as one of the major risks
an
threatening road safety. In this study, we examine in greater detail two important types of traffic
violation events, speeding and drunk driving, and attempt to identify significant risk factors associated
with these types of traffic violations. Risk factors in several different dimensions, including driver,
vehicle, road and environmental factors, are considered. We analyse the speeding (N=11,055) and
M
drunk driving (N=10,035) data for the period 2006-2010 in Guangdong Province, China. These data,
obtained from the Guangdong Provincial Security Department, are extracted from the Traffic
Management Sector-Specific Incident Case Data Report and are the only comprehensive and official
ed

source of traffic accident data in China. Significant risk factors associating with speeding and drunk
driving are identified. We find that several factors are associated with a significantly higher probability
of both speeding and drunk driving, particularly male drivers, private vehicles, the lack of street
pt

lighting at night and poor visibility. The impact of other specific and unique risk factors for either
speeding or drunk driving, such as hukou, road type/grades, commercial vehicles, compulsory third
party insurance and vehicle safety status, also require particular attention. Legislative or regulatory
ce

measures targeting different vehicle types and/or driver groups with respect to the various driver,
vehicle, road and environmental risk factors can subsequently be devised to reduce the speeding and
drunk driving rates. As the country with the highest number of traffic accident fatalities in the world,
Ac

applying these findings in workable legislation and enforcement to reduce speeding and drunk driving
rates will save tens of thousands of lives.

1. Introduction
Speeding- and drunk driving-related traffic fatalities and injuries have been a leading public health
problem worldwide for some time. In most high-income countries, approximately 20% of driver
fatalities are directly related to the driver’s high blood alcohol level at the time of the crash (Global
Road Safety Partnership, 2004a). In New Zealand, 31% of traffic fatalities and 17% of severe injuries


Corresponding author. Tel.: +852 3442 8585; fax: +852 3442 0189.
E-mail address: mskyau@cityu.edu.hk (K.K.W. Yau), zsuzgn@hotmail.com (Guangnan Zhang).

Page 2 of 24
in traffic accidents in the year 2002 were found to be related to speeding (Global Road Safety
Partnership, 2004b). In low- and middle-income countries, the situation is perhaps worse due to the
lack of adequate infrastructure and to inadequate traffic management. Although very little data are
available on the speeding- and drunk driving-related fatalities and injuries, studies have shown that in
low- and middle-income countries, approximately 33% to 69% of drivers killed and 8% to 29% of
drivers injured in traffic accidents had consumed alcoholic beverages prior to driving (Global Road
Safety Partnership, 2004a).

In China (a middle-income country according to World Bank’s country classification by GNI per

t
capita), the number of speeding- and drunk driving-related injuries surged in the years before 2004 and

ip
then began to decline. The Annual Statistical Report on Road Traffic Accidents (Traffic Management
Bureau, Ministry of Public Security) showed that in 2000, a total of 40,873 speeding- and drunk

cr
driving-related injuries were recorded; this number reached its peak in the year 2004, when 73,499
victims were injured. Since then, the amount has fallen steadily; in 2010, the number of speeding- and
drunk driving-related injuries was 57,219, and a 22% reduction overall has been observed since 2004.

us
However, this decrease in number cannot obscure the fact that the magnitude of the decline in speeding
and drunk driving injuries is less significant than that of the reduction of automobile accident-related
injuries. In 2010, the overall number of automobile accident injuries was 254,075, which is nearly half
an
of the 480,865 injuries (reduce by 47%) in 2004. The disproportionate decline between accident-related
injuries and speeding- and drunk driving-related injuries implies that in China, the reduction of
speeding and drunk driving incidents is more challenging compared with other causes of automobile
M
accidents.

For this reason, speeding and drunk driving have become issues of focus for legislative action. On
February 25, 2011, the 8th Amendment to the Criminal Law was passed by the Standing Committee of
ed

the People’s Congress, stipulating that both car-racing on public roads and drunk driving are criminal
offences, whether these actions cause any accidents or not. Under the amended law, whoever drives a
vehicle after drinking shall have his or her license suspended for at least 6 months and pay a minimum
pt

fine of CNY 1,000 (≈USD 160). This amended law has had a positive impact on drunk driving rates in
the short term. Within the four months of its implementation, the number of drunk driving incidents
nationwide decreased by 45.4% compared to the number of incidents recorded in the same period last
ce

year. Whether this legislation continues to be an effective deterrent in the long term remains to be seen.

As a public health issue, speeding and drunk driving have attracted the attention of professionals
Ac

from various sectors. In 2004, a series of reports and manuals published by Global Road Safety
Partnership as well as other organisations, ranging from World Health Organization to the World Bank,
synthesised literature published up to 2004 on the topic of speeding and drunk driving. Apart from
these reports and manuals, a variety of other studies thereafter have also contributed to the study of
these issues from various perspectives.

In general, demographic characteristics are considered to be an important factor affecting speeding

and drunk driving. Among these characteristics, male drivers are usually regarded as a significant risk
factor because they are generally more likely to speed and/or drive drunk than females (Hennessy et al.,
2004; Jonah, 1990; Jonah 1997; OECD, 2006; Smart and Mann, 2002; Wickens et al., 2008, Wickens et
al., 2012; Shinar and Compton, 2004). This is perhaps because males are often overconfident and less
likely to comply with traffic laws when driving, and therefore, they tend to be less cautious about the
perils of dangerous driving behaviours (Shinar and Compton, 2004).

Page 3 of 24
 Likewise, youth is another risk factor that affects speeding and drunk driving. Young drivers are
more likely to speed due to their underestimation of the potential risk of driving situations and
overestimation of their level of skill (Castella and Perez, 2004; Deery, 1999; McKenna and Horswill,
2006; Machin and Sankey, 2008). This risk factor may also be attributed to the inexperience of young
drivers as well as their higher level of excitement-seeking, lower levels of altruism, greater perceived
likelihood of an accident, and lower aversion to risk taking (Machin and Sankey, 2008). Moreover,
young drivers are not only at higher risk of speeding, but studies have also demonstrated that youth is
one of the predictors that profiles repeat drunk driving offenders (Ferguson et al., 1999).

t
Few studies have focused on other demographic factors. However, demographic factors are worthy

ip
of study due to the light they may shed on how individuals’ social backgrounds may affect their driving
behaviour. Of this comparatively small body of literature, Kim et al. (2010), using a random telephone

cr
survey, found that among drivers convicted of drunk driving in Hong Kong, those with an occupation
requiring social drinking are more likely to be involved in traffic accidents. Moreover, women with
higher education levels are found to exhibit a higher likelihood of being involved in drunk driving than

us
their less educated counterparts (Kim et al., 2008; Kim et al., 2010).

In addition to focusing on factors that are associated with demography, several studies have

an
investigated other factors related to vehicles and environment. Speeding has been identified as a
significant factor associated with crashes for commercial vehicles, especially in low-income and
middle-income countries (Global Road Safety Partnership, 2004b). Compared to goods vehicle drivers,
private vehicle drivers have the highest prevalence of a high blood alcohol concentration (BAC), with
M
motorcyclists ranking second in São Paulo. In São Paulo, 6 p.m. to 12 a.m. and 12 a.m. to 6 a.m. were
found to be the time periods with the greatest absolute number of accidents and BAC-positive cases;
Saturdays and Sundays were the days with the highest prevalence of accidents and BAC-positive
ed

accidents (Ponce et al, 2011).

Other research has focused on law enforcement. By investigating the effect of enforcement targeted
at drunk driving in Greece, Yannis et al. (2007) concluded that enforcement can be more successful if
pt

better police practices and resource allocation strategies are adopted. Moreover, when public
campaigns and enforcement targeted at speeding and drunk driving were implemented, they had a
ce

significant independent impact in reducing crashes, while their interactive impact was
anti-complementary; in contrast, although the enforcement and campaigns had no independent impact,
their interactive impact was significant in reducing serious crashes involving young male drivers (Tay,
Ac

2005).

The recent literature generally focuses on the risk factors surrounding drivers, vehicles and
environment, whereas risk factors relating to roads are rarely discussed. Furthermore, in terms of data
sources, few studies are based on data sets extracted from low- and middle-income countries where the
infrastructure, modes of transport and demographic structure are different from those in high-income
countries. As such, this study attempts to determine the significant risk factors associated with (i)
speeding and (ii) drunk driving by analysing traffic accident data in China, with the aim of devising
corresponding traffic regulations and legislative measures targeting different vehicle types and/or driver
groups in terms of the various driver, vehicle and environment risk factors.

Page 4 of 24
2. Materials and Methods

2.1. Data
A recent research study on large scale traffic accident data in China, focusing on traffic violations
and accident severity, established the role of traffic violations as one of the major risks threatening road
safety (Zhang et al., 2013). The current study further examines two important types of traffic violations,
speeding and drunk driving, and attempts to identify significant risk factors associated with these traffic
violations. We analyse the speeding and drunk driving data for the period 2006-2010 in Guangdong

t
Province, China. These data, obtained from the Guangdong Provincial Security Department, are

ip
extracted from the Traffic Management Sector-Specific Incident Case Data Report and are the only
comprehensive and official source of traffic accident data in China recorded by on-scene traffic police.
Together with non-traffic violation accidents, the data contain 11,055 and 10,035 samples relevant to

cr
speeding and drunk driving incidents, respectively. Each sample includes demographic information,
injury severity, vehicle characteristics, road conditions, the time of day of the accident and

us
environmental conditions.

Recent studies on road safety in China mainly rely on observational or survey data, to assess
pedestrian safety (Liu et al., 2011; Zhuang and Wu, 2011), helmet and seatbelt wearing usage (Routley
an
et al., 2008; Huang et al., 2011), and drivers’ attitude and behavior on road safety (Zhang et al., 2006;
Huang et al., 2008). Other studies have adopted regional traffic accident management data (Huang et
al., 2008; Kong et al., 2010), yet not on a large scale. However, research study in China using large
M
scale official database targeting different traffic violation types is rarely found.

Guangdong province is located in the Southern part of Mainland China. Since the reform and
opening-up policy in 1978, its GDP recorded an annual double-digit growth rate every year, which
ed

continuously ranked first among all provinces in the country. As of 2010, the resident population was
104.30 million, the only province in China having a total resident population of over 100 million.
Accident analysis using Guangdong province data is representative. First, due to Guangdong’s rapid
pt

economic development and consequently the vehicle numbers growth, traffic accident incidences are
the highest among all 31 provinces in China. In 2010, the recorded traffic accidents, deaths and injuries
were 13.84%, 9.51% and 14.37% (relative to nationwide total) respectively. Second, Guangdong is
ce

having the highest percent (30%) of floating population (over 36 million) in China. With such a
population mix in the province, risk factors are more representative when compared with the data in
other provinces. This is particularly important when assessing the effect of risk factors relating to
Ac

floating population characteristic such as hukou, occupation, overload and insurance. The
generalization of the research findings to other provinces in China and even to countries outside China
becomes more probable. Third, Guangdong continues to take up the role of national reform
"experimental field" (The National Development and Reform Commission, The Outline of the Plan for
the Reform and Development of the Pearl River Delta (2008-2020), December 2008). Therefore,
current study on Guangdong’s road safety and its corresponding policy recommendations will form
the basis for the development of nationwide traffic management system. So, results arising from
which are exemplary.

To determine the risk factors associated with speeding or drunk driving, the dependent variables are
set as follows: for the presence of speeding or drunk driving, '1' = yes and '0' =no. As the outcome
measure is either 0 or 1, a logistic regression model can be used to estimate the effect of the risk factors

Page 5 of 24
on the probability of (i) speeding or (ii) drunk driving. The risk factors that were considered are the
following: driver, vehicle, road and environment.

2.2. Risk Factors

The risk factors under consideration in the current study have been described in a previous study
derived from the same database (Zhang et al., 2013).

2.2.1. Driver factor

t
Drivers’ age and gender are considered to be potential risk factors. According to the typical driving

ip
habits in different age groups and level of driving skill, driver age is divided into five groups: ≤25,
26-35, 36-45, 46-55 and ≥56. Driving experience is categorised in six groups (measured in years of
experience): ≤2, 3-5, 6-10, 11-15, 16-20 and ≥21.

cr
Although drivers’ education, income and social status are expected to be potential factors that are
associated with traffic violations, this information is generally not recorded in the traffic accident

us
database. Instead, related information indicated by the drivers’ hukou origin and occupation are
available. Two groups of the hukou household registration system are examined: rural hukou and urban
hukou. The rural and urban hukou classification broadly reflects the latent effect of education level,
an
income and social status differences. Urban residents received state-allocated jobs and access to an
array of social services while rural residents were expected to be more self-reliant. Such system
minimizes the movement of people between rural and urban regions, particularly discouraging rural
M
residents moving to cities. However, with the economic growth in China in recent years, huge number
of workers moves from country side to cities to find job. Occupations are classified as military and
police, general staffs, workers, civil servants, the self-employed, farmers, migrant workers, the
ed

unemployed and other occupations.

2.2.2. Vehicle factor

pt

Three major vehicle types, namely, private vehicles, goods vehicles and motorcycles, are included in
the data. In addition, indicator variables are generated according to the vehicle’s safety status, overload
condition, whether the vehicle had any compulsory third party insurance and whether the vehicle was a
ce

commercial vehicle.

2.2.3. Road factor

Ac

Types of traffic lanes can be divided into vehicle lanes, shared lanes and other lanes in accordance
with their functional classification. When considering types and grades, roads can be divided into
expressways, ordinary highways and urban highways. Specifically, ordinary highways include the first,
second, third, fourth and fifth-class highways; urban highways include urban expressways, urban
ordinary highways and urban other highways (detailed road grades classification can be found: GA
17.1-2003 road traffic accident site code Part 1 road type code). There exist big differences between the
road/highway quality in China. Having such a detailed road/highway classification provides accurate
measure of the quality of the road infrastructure and strong discrimination power on the analysis of
speeding and drunk-driving incidences.

Page 6 of 24
2.2.4. Environment factor
There are eight environmental factors under consideration: street-light condition, weather conditions,
visibility level, day of the week, whether it was a public holiday, the time of day, the season and the
year of the accident. Street-light condition is classified into daylight, good street lighting at night and
no street lighting at night. Weather conditions (good=0, bad=1) and visibility level (good=0, bad=1) are
generated as indicator variables accordingly. Weekends are defined as 17:00 Friday to 23:59 Sunday, as
it is anticipated that the occurrence of Friday traffic accidents after 17:00 is similar to those occurring
on Saturday and Sunday (MacLeod, 2011). Public holidays refer to the holidays stipulated by the State

t
Council of China (before 2007, public holidays in China included New Year, Chinese New Year,

ip
International Labour Day and National Day; from 2008 onwards, three holidays were added: the Qing
Ming Festival, Dragon Boat Festival and Mid-Autumn Festival). Time of day has been classified into

cr
six groups: 00:00-06:59, 07:00-08:59, 09:00-11:59, 12:00-16:59, 17:00-19:59, and 20:00-23:59. In
accordance with the Chinese Bureau of Meteorology, seasons are defined as spring (March to May),
summer (June to September), autumn (October to November) and winter (December to February).

us
2.3. Statistical data analysis
Contingency tables are constructed to assess the association between risk factors and (i) speeding and
an
(ii) drunk driving. Chi-square tests of independence are conducted with the level of significance at 5%.
To identify the differences among the risk classes in each factor, corresponding (i) speeding and (ii)
drunk-driving proportion by each risk class within a factor are computed. Such proportion represents
M
the corresponding traffic violation proportion within a particular risk class, which provides insights in
differentiating risk class effect on traffic violation incidences.

To further estimate the effect of different predictor variables on the likelihood of the occurrence of (i)
ed

speeding and (ii) drunk driving, logistic regression analyses are conducted. We consider the use of
multivariate stepwise logistic regressions to identify significant factors determining (i) speeding and (ii)
drunk driving and to estimate the magnitude of adjusted odds ratios (ORs) for each significant factor
pt

while controlling for other confounding factors (Hosmer and Lemeshow, 1989). The adjusted ORs of
significant factors and their 95% confidence intervals (CIs) are computed using a stepwise logistic
regression model in which all factors were initially included and from which insignificant factors were
ce

subsequently removed by the stepwise procedure. Entry and removal probabilities for the stepwise
procedure are both set at 0.05.
Ac

As a remark, results obtained from this study are based on the recorded data of speeding- or drunk
driving-related accidents (that is, non-accident cases are not included in the data). Therefore, the results
on speeding or drunk driving should be interpreted as the effect of risk factors conditional on relevant
accident cases according to the recorded accident data.

3. Empirical results and discussion

3.1. Risk factors affecting speeding

Table 1 presents the chi-square test of independence. As shown in Table 1, a total of 11,055 relevant
car accident cases occurred in 2006-2010 in Guangdong Province, China. The results indicate that both
driver’s gender and age (in years) are important factors significantly associated with speeding. From

Page 7 of 24
Table 2, speeding drivers are more likely to be young (speeding proportion: ‘≤25’= 14.27%,
‘26-35’=13.69%, ’36-45’=14.44%, ’46-55’=11.95%, ‘≥56’=6.85%) and male (speeding proportion:
‘female’=6.95%, ‘male’=13.80%). Additionally, an increased risk of speeding is associated with rural
hukou (speeding proportion: ‘urban hukou’=12.87%, ‘rural hukou’=14.62%), farmers and general staffs
(speeding proportion: ‘farmers’=19.37%, ‘military and police’=15.79%, ‘general staffs’=17.83%,
‘workers’=10.36%, ‘civil servants’=14.88%, ‘self-employed’=14.27%, ‘migrant workers’=12.15%,
‘unemployed’=8.73%, ‘other occupations’=11.44%).

For vehicle factors, vehicle type, safety status, the presence of compulsory third party insurance and

t
commercial operation status are found to have significant relationships with speeding. Specifically,

ip
private vehicles (speeding proportion: ‘motor cycles’=6.58%, ‘private vehicles’=18.14%, ‘goods
vehicles’=15.69%), unfit status (speeding proportion: ‘fit status’=12.97%, ‘unfit status’=23.61%),

cr
compulsory third party insurance (speeding proportion: ‘no’=10.59%, ‘yes’=13.69%) and commercial
operation status (speeding proportion: ‘no’=11.56%, ‘yes’=17.60%) are factors that are all likely to
positively affect speeding.

us
The results indicate that second and third class highways significantly contribute to speeding
(speeding proportion: ‘expressways’=5.78%, ‘first class highways’=14.01%, ‘second class
highways’=22.63%, ‘third class highways’=20.23%, ‘fourth class highways’=18.15%, ‘fifth class
an
highways’=14.71%, ‘urban expressways’=11.87%, ‘urban ordinary highways’=6.89%, ‘urban other
highways’=9.55%).
M
Among environmental factors, it is found that street-light condition, visibility level, public holidays
and time of day exhibit a certain significance. In particular, the highest probability of speeding is found
for no street-lighting at night (speeding proportion: ‘daylight’=13.40%, ‘good street-lighting at
night’=9.96%, ‘no street-lighting at night’=18.65%), poor visibility (speeding proportion: ‘good
ed

visibility’=12.73%, ‘poor visibility’=19.36%), public holidays (speeding proportion: ‘no’=13.15%,

‘yes’=15.75%) and during the time interval 7:00-8:59 (speeding proportion: ‘00:00-6:59’=11.34%,
‘7:00-8:59’=15.67%, ‘9:00-11:59’=13.02%, ‘12:00-16:59’=12.98%,
pt

‘17:00-19:59’=15.13%, ’20:00-23:59’=13.42%).

Coefficients for a logistic regression model estimating the impact of risk factors on speeding are
ce

presented in Table 3, where 10 factors are found to be significantly associating with speeding.
Specifically, an increased probability of speeding is related to male (OR=1.411, 95% CI=1.043-1.908)
and novice drivers (‘≤2’: OR=1.501, 95% CI=1.268-1.777; ‘3-5’: 1.277, 95% CI=1.084-1.503; ‘6-10’:
Ac

OR=1.168, 95% CI=1.004-1.358). Among all occupations, farmers/military and police/general

staffs/civil servants all exhibit higher probabilities of speeding.

Among vehicle factors, goods vehicles, private vehicles (private vehicles: OR=3.886, 95%
CI=3.307-4.566; goods vehicles: OR=2.687, 95% CI=2.201-3.280), unfit status (OR=2.087, 95%
CI=1.606-2.713) and commercial operation status (OR=1.232, 95% CI=1.063-1.429) contributed
significantly to the occurrence of speeding. Expressways exhibit a particularly low risk of speeding
compared to other highway types. From Table 3, street-light conditions, visibility level, the time of day
and years are identified to be significant factors affecting speeding. Specifically, no street-lighting at
night (OR=1.596, 95% CI=1.358-1.875) demonstrates an increasing probability of speeding, as does
poor visibility (OR=1.314, 95% CI=1.092-1.582) and the time of day 07:00-08:59 (OR=1.312, 95%
CI=1.075-1.601). However, 00:00-06:59 (OR=0.750, 95% CI=0.632-0.889) and the year 2008

Page 8 of 24
(OR=0.835, 95% CI=0.722-0.967) are shown to have a decreased speeding probability.

Factors without any significant association with speeding include the driver’s age, vehicle overload
condition, compulsory third party insurance, type of traffic lanes, day of the week, public holidays and
seasons.

3.2. Risk factors affecting drunk driving

Similar to the analysis of risk factors determining speeding, a total of 10,035 relevant car accident
cases were examined. Table 1 presents that among the factors describing drivers’ characteristics, the

t
driver’s gender, age, hukou origin and occupation exhibit significant associations with drunk driving.

ip
Specifically, from Table 2, male drivers (drunk driving proportion: ‘female’=1.14%, ‘male’=4.78%),
young drivers (drunk driving proportion: ‘≤25’= 5.61%, ‘26-35’=5.10%, ‘36-45’=4.16%,

cr
‘46-55’=2.30%, ‘≥56’=4.94%), and urban hukou registrants (drunk driving proportion: 'urban
hukou'=5.06%, 'rural hukou'=2.96%) are more likely to become involved in drunk driving. Among the
occupations, military and police present the highest risk of drunk driving, while farmers have the

us
lowest risk (drunk driving proportion: ‘farmers’=3.22%, ‘military and police’=8.57%, ‘general
staffs’=5.18%, ‘workers’=3.52%, ‘civil servants’=6.54%, ‘self-employed’=7.15%, ‘migrant
workers’=4.61%, ‘unemployed’=7.93%, ‘others’=3.57%).
an
The results also indicate that drunk driving incidents are more likely to occur in private vehicles
(drunk driving proportion: ‘motor cycles’=3.76%, ‘private vehicles’=7.22%, ‘goods vehicles’=1.81%),
and in conditions where vehicles are not overloaded (drunk driving proportion: ‘not overload’=4.72%,
M
‘overload’=1.72%), had no compulsory third party insurance (drunk driving proportion: ‘no’=6.29%,
‘yes’=4.28%) or are not commercial operation vehicles (drunk driving proportion: ‘no’=5.74%,
‘yes’=1.22%).
ed

For road factors, only two road types demonstrate significant association with drunk driving. Table 2
shows that a majority of drunk driving incidents happen on urban ordinary highways and urban other
highways (drunk driving proportion: ‘expressways’=0.28%, ‘first class highways’=3.63%, ‘second
pt

class highways’=2.44%, ‘third class highways’=5.70%, ‘fourth class highways’=5.51%, ‘fifth class
highways’=4.57%, ‘urban expressways’=5.59%, ‘urban ordinary highways’=6.74%, 'urban other
ce

highways'=5.91%).

Considering environmental factors, both street-light condition and time of the accident demonstrate a
strong association with drunk driving incidents. Specifically, good street-lighting at night (drunk
Ac

driving proportion: ‘daylight’=0.93%, ‘good street-lighting’=11.61%, ‘no street-lighting’=4.04%) and

the times 20:00-23:59 and 00:00-6:59 (drunk driving proportion: ‘00:00-6:59’=10.59%,
‘7:00-8:59’=0.88%, ‘9:00-11:59’=0.28%, ‘12:00-16:59’=1.46%, ‘17:00-19:59’=2.21%,
‘20:00-23:59’=9.19%) have higher shares of drunk driving cases. Poor visibility (drunk driving
proportion: ‘good visibility’=3.92%, ‘poor visibility’=10.57%), weekends (drunk driving proportion:
‘weekdays’=4.06%, ‘weekends’=5.55%), public holidays (drunk driving proportion: ‘no’=4.33%,
‘yes’=7.06%) and recent years (2009 and 2010) (drunk driving proportion: ‘2006’=3.19%,
‘2007’=3.73%, ‘2008’=4.45%, ‘2009’=6.65%, ‘2010’=5.36%) also display higher than average
proportions of drunk driving incidents. However, the driver’s driving experience, vehicle safety status,
type of vehicle lanes and seasons show no significant association with drunk driving.

By applying stepwise logistic regression analysis, results show that a higher risk of drunk driving is

Page 9 of 24
more likely to be associated with male drivers (OR=4.849, 95% CI=2.352-9.995) and a lower risk in
drunk driving with novice drivers (OR=0.740, 95% CI=0.563-0.972), rural hukou registrants
(OR=0.664, 95% CI=0.504-0.874) and 46-55 year- olds (OR=0.505, 95% CI=0.331-0.772).

Table 3 also indicates that private vehicles (OR=1.856, 95% CI=1.500-2.296) exhibit an increasing
likelihood to become involved in drunk driving, whereas compulsory third party insurance (OR=0.712,
95% CI=0.536-0.945) is negatively associated with drunk driving. It is worth emphasising that
commercial operation vehicles (OR=0.268, 95% CI=0.184-0.390) present a significantly lower
probability of drunk driving. For road factors, drunk driving occurs more frequently in shared lanes

t
(OR=1.560, 95% CI=1.195-2.036).

ip
Relative to daylight, both good street lighting at night (OR=14.182, 95% CI=6.416-31,350) and no
street-lighting at night (OR=7.537, 95% CI=3.280-17.318) exhibit a higher probability to become

cr
scenes of drunk driving incidents, as do areas with poor visibility (OR=1.812, 95% CI=1.387-2.368)
and public holidays (OR=1.574, 95% CI=1.128-2.198). When considering the time factor, drunk

us
driving incidents are more likely to happen between 00:00-06:59 (OR=4.857, 95% CI=3.288-7.173),
07:00-08:59 (OR=3.724, 95% CI=1.290-10.755), 12:00-16:59 (OR=5.760, 95% CI=2.542-13.053) and
20:00-23:59 (OR=2.583, 95% CI=1.743-3.827). Moreover, both 2009 and 2010 demonstrate higher

an
rates of drunk driving (2009: OR=1.809, 95% CI=1.410-2.320; 2010: OR=1.635, 95%
CI=1.248-2.141), with the numbers of drunk driving cases increasing significantly since 2008.

Conversely, vehicle safety status, overload condition, day of the week and seasons were not
M
identified as significant factors affecting drunk driving by stepwise logistic regression analysis.

3.3. Discussion of results

Consistent with studies extant in the literature, we found that male drivers are more likely to become
ed

involved in speeding and drunk driving incidents (Hennessy et al., 2004; Jonah, 1990; Jonah, 1997;
OECD, 2006; Shinar and Compton, 2004; Smart and Mann, 2002; Wickens et al., 2008; Wickens et al.,
2012), as shown in Table 3. In general, female drivers are more cautious and are less aggressive than
pt

male drivers in driving attitude. This problem is indeed serious because the majority of drivers in China
are male: the National Economic and Social Development Statistics Bulletin released by National
ce

Statistics Bureau of China demonstrated that until 2011, of the 230 million licensed drivers in China,
approximately 81.04% were male. It should also be noted that in China, the gender discrepancy in
drunk driving is even more significant than that of speeding.
Ac

Private vehicles exhibit a significant association with both speeding and drunk driving. Such finding
generally agrees with the results obtained in a recent study (Ponce et al., 2011). In addition, speeding
and drunk driving cases are more likely to take place during the morning rush hour (07:00-08:59) and
under conditions where there is no street-lighting at night and poor visibility levels. When the visibility
level is poor or there is no street lighting at night, it is possible that police enforcement targeted at
speeding and drunk driving would be less intense and that vehicles and pedestrians would be less dense
due to the low visibility. This perhaps indicates that at night time, the intensification of police
enforcement and improved visibility conditions have negative associations with speeding and drunk
driving rates. Thus, the improvement of lighting facilities on those streets where street lighting does not
exist or does not function may help to reduce speeding and drunk driving incidents. Random police
enforcement patrols at night and in low visibility areas may also help to reduce these incidents.

Page 10 of 24
 Novice drivers (≤2 years of driving experience) are more likely to speed but less likely to drive
drunk. Similarly, commercial operation vehicles exhibit higher rates of speeding (Global Road Safety
Partnership, 2004b), yet their risk of drunk driving is particularly low. In addition, drunk driving cases
are more likely to take place during the evening and at dawn (20:00-23:59 and 00:00-06:59) (Ponce et
al., 2011).

Our results show that certain occupation groups, such as workers and migrant workers, exhibit
significantly lower rates of speeding and drunk driving. Related research study also found the
association between occupation category and drunk driving behaviour (Kim et al., 2010). Expressways

t
show particularly low rates of speeding and drunk driving. Because expressways are high grade

ip
highways suitable for high speed driving, drivers may be more cautious when driving on expressways
due to safety considerations.

cr
A driver’s level of driving experience is only a significant factor in speeding: as a driver’s driving
experience grows, speeding rates decrease. In addition, third and fourth class highways are the road

us
types that have higher rates of speeding, in contrast with urban ordinary highways, which exhibit the
lowest risk of speeding. Compared to motorcycles, we found that private vehicles and goods vehicles
exhibit a higher probability of speeding. Additionally, vehicles with unfit safety statuses are more likely

an
to speed. Moreover, in the year 2008, self-employed and unemployed drivers were also found to have a
lower risk of speeding, yet their risk of drunk driving is not significant.

The highest risks of drunk driving are found for 46-55 year-olds and rural hukou registrants.
M
However, in other studies, young drivers were found more likely to involve in risk taking driving
behaviour (Ferguson et al., 1999; Machin and Sankey, 2008). Drunk driving is less likely to occur when
the vehicle has compulsory third party insurance, and drunk driving cases are more likely to happen in
shared-lanes. An increased incidence of drunk driving incidents is also observed during public holidays.
ed

Moreover, non-peak day time (12:00-16:59) is identified to be a significant factor affecting drunk
driving, and the years 2009 and 2010 show particularly high rates of drunk driving incidents. This may
be attributed to the fact that the volume of vehicles in Guangdong Province, China, greatly surged
pt

during 2009 and 2010 as a result of the CNY 4 trillion governmental bailout in response to the 2008
financial turmoil. In 2009, according to the Yearbook of China – Transportation & Communication,
ce

1,904,764 new vehicles were registered, compared with the 1,563,050 and 961,149 new registered
vehicles in 2008 and 2007, respectively; the 2009 figures account for approximately 15% of the total
vehicle amount in the same year. Moreover, in 2010, the amount of new registered vehicles was
Ac

2,638,724, nearly 34% more than the amount of annual new registered vehicles in 2009 and accounting
for almost 17% of the total number of vehicles in Guangdong Province. Notably, even the drunk
driving rate has stayed constant; however, the fast growing volume of vehicles could still have a
significant impact on the number of drunk driving cases in 2009 and 2010. Overload condition, day of
the week and seasons are not found to be significant factors associated with speeding and drunk driving.
Whilst in another study, Saturdays and Sundays were the days with the highest prevalence of accidents
and BAC-positive accidents (Ponce et al., 2011).

4. Conclusions and recommendations

With respect to theoretical contribution to the science literature, firstly this study employed a large
dataset as its sample. This dataset includes and focuses on potential risk factors that the existing

Page 11 of 24
literature may not have considered. Thus, specific risk factors in the context of China may be
investigated and assessed, in addition to risk factors commonly chosen in the literature. Because these
distinct risk factors were included in the logistic regression models, unexplained variability in the error
component can be reduced, and the results are more accurate and reliable. The following risk factors
merit special attention: hukou (urban-rural distinction), occupation categories and road type/grades.

Hukou was identified to be a significant factor that is negatively associated with drunk driving (for
rural hukou: OR=0.664 and 95% CI=0.504-0.874). In general, whereas rural hukou holders are mostly
farmers, urban hukou holders are more likely to have occupations requiring frequent business dinners.

t
Additionally, drinking is required as part of the etiquette to foster relations among businessmen during

ip
such business dinners in China, which leads to an increase in the probability that urban hukou drivers
will engage in drunk driving.

cr
The results also reveal that the occupation of the driver significantly affects driving behavior in
China. For example, using farmers for comparison, the odds ratios (95% CI) of speeding and drunk

us
driving for migrant workers are 0.616 (0.492-0.772) and 0.624 (0.437-0.892), respectively. For workers,
the corresponding figures are 0.516 (0.434-0.614) and 0.615 (0.464-0.813), respectively. The level of
education of farmers tends to be relatively low, which may mean that farmers may be more likely to

an
disobey traffic regulations. However, migrant workers—farmers that have emigrated from rural areas
to urban areas and work mostly in factories—are less likely to be involved in speeding and drunk
driving than farmers. Furthermore, workers are even less likely to be involved in speeding and drunk
driving. In China, workers typically enjoy better socio-economic status and a higher level of education
M
than farmers and migrant workers. This demonstrates that drivers whose occupations require a higher
level of education and have better socio-economic status will rationally choose not to violate traffic
regulations.
ed

A detailed classification of road type/grades (expressways, first-class highways, second-class

highways, third-class highways, fourth-class highways, fifth-class highways, urban expressways,
ordinary urban highways, and other urban highways) offers a clear distinction for various road qualities
pt

and an accurate estimation of the effects of differently leveled road infrastructures on speeding and
drunk driving. For example, traffic violations are more likely to occur on ordinary urban highways than
ce

on expressways. In general, expressways are better equipped with speed monitoring devices. When
driving on expressways, drivers understand that there is greater risk of being caught and facing a
substantial penalty if they violate traffic regulations. In addition, expressway drivers are more likely to
Ac

be long-distance travelers that are cognizant of the potential dangers of traffic violations and are
therefore more vigilant. For example, alcohol significantly harms the response abilities of drivers;
experienced drivers thus prefer not to drink before driving on expressways. Our findings also indicate
that traffic violations are less likely to occur on urban expressways for similar reasons. Traffic
enforcement and speed monitoring devices are more intensely used on these types of roads. In addition,
there is frequently traffic congestion on urban expressways, and speeding is less likely in such a
crowded environment. Poor road quality (e.g., fifth-class highways) also implies less speeding; when
the road surface conditions are poor, drivers are reluctant to speed. The results for first-class highways,
second-class highways, third-class highways, fourth-class highways and fifth-class highways broadly
indicate a compromise between traffic police enforcement intensity and decreasing road quality. Higher
traffic police enforcement intensity and lower road quality lead to fewer (recorded) speeding violations.

Secondly, this study assessed speeding and drunk driving incidents separately instead of

Page 12 of 24
combining them in the data analysis (differing from certain previous studies); thus, unique risk factors
associated with either speeding or drunk driving can be determined separately from one another. The
following list highlights certain significant risk factors: driving experience, commercial vehicles,
compulsory third party insurance, vehicle safety status and visibility.

Although it is generally expected that novice drivers (less than 2 years driving experience) should
be more likely to be associated with traffic violations, the current study has shown that novice drivers
are more likely to be involved in speeding violations (odds ratio=1.501; 95% CI=1.268-1.777),
whereas they are less likely to be involved in drunk driving violations (odds ratio=0.740; 95%

t
CI=0.563-0.972). It is notable that similar pattern on speeding and drunk driving is found for

ip
commercial vehicles. In particular, commercial vehicles are more likely to be involved in speeding
(odds ratio=1.232; 95% CI=1.063-1.429) but are less likely to be involved in drunk driving (odds

cr
ratio=0.268; 95% CI=0.184-0.390). Such findings would not be revealed if speeding and drunk driving
data were combined, which might lead to risk factor effect cancelation.

us
According to the moral hazard principle, drivers with vehicle insurance policies are expected to be
less cautious when driving because the potential costs or burdens of risky behavior will be carried (in
whole or in part) by the insurance company, which leads to the greater likelihood that insured drivers

an
will commit speeding and drunk driving violations. However, our study indicated that insured drivers
have a significantly lower risk of being involved in drunk driving (odds ratio=0.712; 95%
CI=0.536-0.945), whereas insurance's association with speeding is not significant. Moreover, vehicles
with a good safety status were expected to be more likely to be involved in speeding. This study has
M
found that vehicles not in good safety status, however, were more likely to be involved in speeding
(odds ratio=2.087; 95% CI=1.606-2.713). Bad visibility was also positively associated with both
speeding (odds ratio=1.314; 95% CI=1.092-1.582) and drunk driving (odds ratio=1.812; 95%
ed

CI=1.387-2.368). This finding was unexpected because drivers were expected to drive more carefully
when visibility was poor. In China, however, the reverse is true. This may be the result of less intensive
police enforcement when there is poor visibility; drivers may perceive low probability of being caught
pt

in violations and behave more recklessly.

This study has established that speeding- and drunk driving-related traffic accidents are growing
ce

public health problems in China. Over the past several years, a set of measures has successfully
reduced the speeding and drunk driving rates elsewhere, and furthermore, the impact of road traffic
injuries (WHO, 2004). Similar targeted measures and interventions can be implemented in China to
Ac

reduce speeding and drunk driving incidents. In the literature, it is established that when designing road
traffic interventions to reduce traffic violation incidences, measures pinpointing to drivers’ attitude
played a critical role (Forward, 2009; Poulter et al., 2008; Machin and Sankey, 2008). In particular, for
speeding and drunk driving events, it is well established that a change in the driver’s attitude is of
utmost importance in order to decrease speeding (Tranter and Warn, 2008; Elliott et al., 2005) and
drunk driving (Jornet-Gibert et al., 2005) incidences. To be efficient in these intervention programs,
targeted groups of drivers should be emphasized, such as young drivers (Forward, 2009) and truck
drivers (Poulter et al., 2008). To maximize the potential for reducing drunk driving incidence, some
studies suggested to take into consideration the effect of gender (Butters et al., 2012) and social context
(Beck et al., 2011). Also, a dramatic increase in enforcement including an increment in social
marketing tactics by changing the belief and making drunk driving unacceptable can reduce drunk
driving events (Dula et al., 2007). While interventions focusing on sensation seekers and environmental

Page 13 of 24
factors that mediate the link between sensation seeking and alcohol-impaired driving behaviors will
help reducing speeding incidence (Zakletskaia et al., 2009). Specific countermeasures by integrating
interventions into driving education program with focus on personal psychological risk factors are
found to be effective in improving traffic safety (Paaver et al., 2012). In addition, a multidisciplinary
approach including public education, stepped-up enforcement, and psychological treatments is
proposed (Fakhry and Salaita, 2002). Besides, Tran et al. (2012) called for a comprehensive approach,
including social marketing and public education campaigns, enhanced enforcement, and programs.

As commercial vehicles exhibit a higher likelihood of being involved in speeding incidents,

t
chain-of-responsibility principles could be introduced. The chain of responsibility principle states that

ip
any individual or organisation that has influence in the transport chain has the responsibility to ensure
compliance with the law. If these principles are included in the law, employer influence may be able to

cr
reduce speeding incidents for commercial vehicle drivers. Vehicles with unfit safety statuses were
demonstrated to be significantly associated with speeding. In spite of intensified police enforcement
focused on speeding and more rigorous annual vehicle inspections, countermeasures should also

us
include action aimed at boosting vehicle scrap rate. According to estimates from the China Automobile
Dealers Association, the annual scrap rate in China is 7%, which is equivalent to approximately 2
million motor vehicles, of which only about one fourth are actually junked (Beijing Traffic
an
Management Bureau, 2010). Thus, a used vehicle recycling law should be legislated in China to
regulate the vehicle recycling market, reduce government intervention in the management of scrapped
vehicles, and motivate vehicle owners to junk their scrapped vehicles.
M
In China, a drunk driver with a BAC between 0.02 and 0.08 is given a 6 month license suspension,
while some studies have concluded that a longer license suspension, e.g., 12 to 18 months, is more
efficient than a shorter suspension (Ferguson et al, 1999). Thus, longer length suspensions could be
ed

employed to better deter potential drunk drivers. Business dinner- and banquet-related drinking is
inevitable for business in China; moreover, drinking is a necessary part of the etiquette to foster
relations among businessmen. This combination increases the probability of drunk driving for the
pt

participants of dinners and banquets. In this case, more stringent police enforcement could be
implemented around high risk locations, such as large hotels and restaurants, to prevent inebriated
businessmen from driving. Such enforcement would be more effective, however, if it is supported by
ce

intensive publicity campaigns, especially television advertisements, to promote designated drivers.

The results obtained in this study may likely be generalized to other provinces in China (particularly
Ac

those risk factors that are specific to China) because the accident data (in Guangdong Province) and
risk factors considered are rather representative and comprehensive. The fact that Guangdong has a
high percentage of residents migrating from other provinces, and is having one of the largest number of
vehicles among all provinces in China reinforces the generalizability. Especially, since we consider the
problems with respect to traffic safety in Guangdong which is typical in reflecting the problems faced
by other provinces, our findings can be a direct guideline for governments to design traffic safety
policy at a national level. For example, existing literatures generally agreed that socio-economic status
is a significant factor influencing traffic accident incidence. Moreover, the fatality rate with respect to
traffic accidents shows distinct diversity among social classes: people with low social status or living in
poor areas may suffer from the greatest probability to become involved in traffic injuries and fatalities
(WHO, 2004). This conclusion was further confirmed by current research findings as farmers, workers
and migrant workers were identified to be the occupations with the highest likelihood to become

Page 14 of 24
involved in speeding and drunk driving. Such a finding suggested that policy designers should be
aware that different risks of speeding and drunk driving were relating to different occupation groups.

Moveover, since the classification of hukou as well as the road category is the same across provinces,
our results are of reference value for other provinces. Such as: there is a higher possibility for people
with urban hukou to be involved in drunk driving compared to their rural counterparts; drivers are less
likely to commit speeding and drunk driving when driving on expressways than on ordinary urban
highways. These findings relating to hukou and road factors may contribute to the designing of efficient
road policies in other provinces of China. For regions outside China, the effects of unique risk factors

t
associated with speeding and drunk driving remain to be verified. Nevertheless, our findings provide a

ip
more intuitive and reliable theoretical and analytical framework from which to formulate policy that
may also be useful to low- and middle-income countries.

cr
In conclusion, speeding- and drunk driving-related traffic accidents have been a leading cause of
fatalities and injuries in China. By examining the risk factors that determine speeding and drunk

us
driving, this study identified that male drivers, private vehicles, a lack of street lighting at night and
poor visibility are factors demonstrating significantly a higher probability of both speeding and drunk
driving incidents. The impact of other specific and unique risk factors for either speeding or drunk

an
driving, such as hukou, road type/grades, commercial vehicles, compulsory third party insurance and
vehicle safety status, also require particular attention. As the country with the highest number of traffic
accident fatalities in the world, applying these findings in workable legislation and enforcement to
reduce speeding and drunk driving rates will save tens of thousands of lives.
M
Acknowledgements
ed

The authors are grateful to an Associate Editor and two reviewers for very helpful comments on
earlier versions of the paper. This research was supported in part by the National Natural Science
Foundation of China grant 71173244. We would also like to thank Haoping Yu of China Guangfa Bank
pt

for his constructive suggestions and advice.

ce
Ac

Page 15 of 24
Table 1
Chi-square Test of Independence
Factors Speeding (N=11,055) Drunk-driving (N=10,035)
χ2 (P-value), χ2 (P-value),
degree of freedom degree of freedom
Driver factors
Driver’s gender 28.273*(<0.001),1 20.117*(<0.001),1
Driver’s age 12.748* (0.013),4 20.899 * (<0.001),4
Driver’s driving experience

t
5.133 (0.400),5 8.277 (0.142),5

ip
Driver’s hukou origin 5.672* (0.017),1 19.639*(<0.001),1
Driver’s occupation 98.229*(<0.001),11 59.146*(<0.001),11

cr
Vehicle factors
Vehicle type 253.693*(<0.001),2 109.786*(<0.001),2
Vehicle safety status 35.654*(<0.001),1 0.510(0.475),1

us
Overload condition 1.850(0.174),1 12.456*(<0.001),1
Compulsory third party insurance 9.448* (0.002),1 10.024* (0.002),1
Commercial operation 72.258*(<0.001),1 93.3010*(<0.001),1
Road factors
Type of vehicle lane
an3.693(0.055),1 3.424 (0.064),1
Road type 406.637*(<0.001),11 110.737*(<0.001),11
M
Environmental factors
Street-light condition 71.402*(<0.001),2 511.208*(<0.001),2
Weather condition 0.601(0.438),1 6.276* (0.012),1
Visibility level 34.875*(<0.001),1 84.556*(<0.001),1
ed

Day of the week 2.651(0.104),1 10.993* (0.001),1

Public holiday 4.244* (0.039),1 11.359* (0.001),1
Time 16.063*(0.007),5 413.846*(<0.001),5
pt

Seasons 2.154(0.541),3 5.054(0.168),3

Years 5.613(0.230),4 33.748*(<0.001),4
ce

*Significant at 5% level.
Ac

Page 16 of 24
Table 2
Speeding and Drunk-driving Proportion by Risk Factor Categories
Speeding Drunk-driving
Driver’s gender
Female 6.95 1.14
Male 13.80 4.78
Driver’s age
14.27 5.61
≦25

t
13.69 5.10

ip
26-35 13.44 4.16
36-45 11.95 2.30
46-55 6.85 4.94

cr
≧56

us
Driver’s driving experience
14.58 4.50
≦2
13.13 5.26
3-5
6-10
11-15
an 12.96
12.52
14.20
4.59
3.64
5.12
M
16-20 13.26 3.58

≧21

Driver’s hukou origin

ed

Urban hukou 12.87 5.06

Rural hukou 14.62 2.96
Driver’s occupation
pt

Farmers 19.37 3.22

Military and police 15.79 8.57
ce

General staffs 17.83 5.18

Workers 10.36 3.52
Civil servants 14.88 6.54
Ac

Self-employed 14.27 7.15

Migrant workers 12.15 4.61
Unemployed 8.73 7.93
Other occupations 11.44 3.57
Vehicle type
Motor cycle 6.58 3.76
Private vehicle 18.14 7.22
Goods vehicle 15.69 1.81
Vehicle safety status
Fit status 12.97 4.55
Unfit status 23.61 3.68
Overload condition

Page 17 of 24
 Not overload 13.45 4.72
Overload 11.66 1.72
Compulsory third party insurance
No 10.59 6.29
Yes 13.69 4.28
Commercial operation
No 11.56 5.74
Yes 17.60 1.22
Type of vehicle lanes

t
ip
Shared-lanes 14.72 5.38
Other lanes 13.05 4.35
Road type

cr
Expressways 5.78 0.28
The first class highways 14.01 3.63

us
The second class highways 22.63 2.44
The third class highways 20.23 5.70
The fourth class highways 18.15 5.51
The fifth class highways
Urban expressways
Urban ordinary highways
an
14.71
11.87
6.89
4.57
5.59
6.74
Urban other highways 9.55 5.91
M
Street-light condition
Daylight 13.40 0.93
Good street-lighting at night 9.96 11.61
ed

No street-lighting at night 18.65 4.04

Weather condition
Good weather 13.48 4.22
pt

Bad weather 12.89 5.42

Visibility level
Good visibility 12.73 3.92
ce

Bad visibility 19.36 10.57

Day of the week
Weekdays 12.98 4.06
Ac

Weekends 14.12 5.55

Public holidays
No 13.15 4.33
Yes 15.75 7.06
Time
00:00-06:59 11.34 10.59
07:00-08:59 15.67 0.88
09:00-11:59 13.02 0.28
12:00-16:59 12.98 1.46
17:00-19:59 15.13 2.21
20:00-23:59 13.42 9.19

Page 18 of 24
Seasons
Spring 14.03 4.45
Summer 12.81 4.05
Autumn 13.46 4.48
Winter 13.21 5.28
Years
2006 12.82 3.19
2007 13.25 3.73
2008 12.38 4.45

t
ip
2009 14.28 6.65
2010 14.39 5.36

cr
Overall 13.33 4.52

us
an
M
ed
pt
ce
Ac

Page 19 of 24
Table 3
Adjusted Odds Ratios (95% Confidence Intervals) in Stepwise Logistic Regression Analysis
Factors Speeding Drunk-driving
Odds ratio (95% CI) Odds ratio (95% CI)
Driver’s gender (base: female)
Male 1.411(1.043,1.908) 4.849(2.352,9.995)

Driver’s age (base: ≧56)

16-25 - -

t
ip
26-35 - -
36-45 - -
46-55 - 0.505(0.331,0.772)

cr
Driver’s driving experience (base: ≧21)

us
1.501(1.268,1.777) 0.740(0.563,0.972)
≦2

3-5 1.277(1.084,1.503) -
6-10
11-15
16-20
an
1.168(1.004,1.358)
-
-
-
-
-
Driver’s hukou origin (base: urban hukou)
M
Rural hukou - 0.664(0.504,0.874)
Driver’s occupation (base: farmers)
Military and police - -
ed

General staffs - -
Workers 0.516(0.434,0.614) 0.615(0.464,0.813)
Civil servants - -
pt

Self-employed 0.715(0.601,0.851) -
Migrant workers 0.616(0.492,0.772) 0.624(0.437,0.892)
ce

Unemployed 0.481(0.295,0.782) -
Other occupations 0.677(0.576,0.796) 0.559(0.430,0.727)
Vehicle type (base: motor cycle)
Private vehicles 3.886(3.307,4.566) 1.856(1.500,2.296)
Ac

Goods vehicles 2.687(2.201,3.280) -

Vehicle safety status (base: fit status)
Unfit status 2.087(1.606,2.713) -
Overload condition (base: not overload)
Overload - -
Compulsory third party insurance (base: no)
Yes 0.712(0.536,0.945)
Commercial operation (base: no )
Yes 1.232(1.063,1.429) 0.268(0.184,0.390)
Type of vehicle lanes (base: other lanes)
Shared-lanes - 1.560(1.195,2.036)

Page 20 of 24
Road types (base: the first class highways)
Expressways 0.268(0.201,0.358) 0.074(0.023,0.240)
The second class highways 1.819(1.567,2.110) 0.654(0.470,0.911)
The third class highways 1.653(1.356,2.016) -
The fourth class highways 1.510(1.098,2.078) -
The fifth class highways - -
Urban expressways - -
Urban ordinary highways 0.559(0.467,0.670) -
Urban other highways - -

t
ip
Street-light condition (base: daylight)
Good street-lighting at night - 14.182(6.416,31.350)
No street-lighting at night 1.596(1.358,1.875) 7.537(3.280,17.318)

cr
Weather condition (base: good)
Bad weather - -

us
Visibility level (base: good)
Bad visibility 1.314(1.092,1.582) 1.812(1.387,2.368)
Day of the week (base: no)
Weekends
Public holidays (base: no)
Yes
an -

-
-

1.574(1.128,2.198)
M
Time (base: 09:00-11:59)
00:00-06:59 0.750(0.632,0.889) 4.857(3.288,7.173)
07:00-08:59 1.312(1.075,1.601) 3.724(1.290,10.755)
12:00-16:59 - 5.760(2.542,13.053)
ed

17:00-19:59 - -
20:00-23:59 - 2.583(1.743,3.827)
Seasons (base: autumn)
pt

Spring - -
Summer - -
ce

Winter - -
Years (base:2006)
2007 - -
2008 0.835(0.722,0.967) -
Ac

2009 - 1.809(1.410,2.320)
2010 - 1.635(1.248,2.141)

Percent correct 86.649 95.436

ROC 0.736 0.872
N 11,055 10,035

Page 21 of 24
References
Beck, K. H., Ahmed, A., Farkas, Z. A., 2011. A descriptive analysis of the social context of drinking
among first-time DUI offenders. Traffic Injury Prevention 12(4), 306-311.

Beijing Traffic Management Bureau, 2010. More old cars need to be junked. Retrieved November 2012

t
from http://www.bjjtgl.gov.cn/publish/portal1/tab165/info19715.htm.

ip
Butters, J., Mann, R. E., Wickens, C. M., Boase, P., 2012. Gender differences and demographic
influences in perceived concern for driver safety and support for impaired driving countermeasures.

cr
Journal of Safety Research 43, 405-411.

Castella, J., Perez, J., 2004. Sensitivity to punishment and sensitivity to reward and traffic violations.

us
Accident Analysis and Prevention 36, 947–952.

Deery, H. A., 1999. Hazard and risk perception among young novice drivers. Journal of Safety
Research. 30 (4), 225–236.
an
Dula, C. S., Dwyer, W. O., LeVerne, G., 2007. Policing the drunk driver: Measuring law enforcement
involvement in reducing alcohol-impaired driving. Journal of Safety Research 38, 267-272.
M
Elliott, M. A., Armitage, C. J., Baughan, C. J., 2005. Exploring the beliefs underpinning drivers
intentions to comply with speed limits. Transportation Research Part F 8, 459-479.

Fakhry, S. M., Salaita, K., 2002. Aggressive driving: A preliminary analysis of a serious threat to
ed

motorists in a large metropolitan area. Journal of Trauma-Injury Prevention and Critical Care 52(2),
217-223.

Ferguson, M., Sheehan, M., Davey, J., Watson, B., 1999. Drink driving rehabilitation: The present
pt

context.

Forward, S. E., 2009. An assessment of what motivates road violations. Transportation Research Part F
ce

12, 225–234.

Global Road Safety Partnership. 2004a. Drinking and Driving: A road safety manual for
decision-makers and practitioners.
Ac

Global Road Safety Partnership. 2004b. Speed Management: A road safety manual for decision-makers
and practitioners.

Hennessy, D. A., Wiesenthal, D. L., Wickens, C. M., Lustman, M., 2004. The impact of gender and
stress on traffic aggression: Are we really that different? In J. P. Morgan (Ed.), Focus on aggression
research (pp. 157–174). Hauppauge, NY: Nova Science Publishers.

Hosmer, D. W., Lemeshow, S., 1989. Applied Logistic Regression. John Wiley and Sons, New York.

Huang, Y. H., Rau, P. L. P., Zhang, B., Roetting, M., 2008. Chinese truck drivers’ attitudes toward
feedback by technology: A quantitative approach. Accident Analysis and Prevention 40, 1553-1562.

Huang, Y. H., Zhang, W., Morphy, L., Shi, G., Lin, Y., 2011. Attitudes and behavior of Chinese drivers
regarding seatbelt use. Accident Analysis and Prevention 40, 889-897.

Page 22 of 24
Jonah, B. A., 1990. Age differences in risky driving. Health Education Research: Theory & Practice, 5,
139–149.

Jonah, B. A., 1997. Sensation seeking and risky driving: A review and synthesis of the literature.
Accident Analysis and Prevention 29, 651–665.

Jornet-Gibert, M., Gallardo-Pujol, D., Suso, C., Andres-Pueyo, A., 2013. Attitudes do matter: The role
of attitudes and personality in DUI offenders. Accident Analysis and Prevention 50, 445-450.

Kim, J. H., Lee, S., Chan, K. W. C., Lau, J., Tsang, A., Griffiths, S. M., 2010. A population-based study

t
on the prevalence and correlates of drinking and driving in Hong Kong. Accident Analysis and

ip
Prevention 42, 994-1002.

Kim, J. H., Lee, S., Chow, J., Lau, J., Tsang, A., Choi, J., Griffiths, S., 2008. Prevalence and factors

cr
associated with binge drinking, alcohol abuse and alcohol dependence: a population-based study of
Chinese adults in Hong Kong. Alcohol and Alcoholism 43 (3), 360-370.

us
Kong, C., Yang, J., 2010. Logistic regression analysis of pedestrian casualty risk in passenger vehicle
collisions in China. Accident Analysis and Prevention 42, 987-993.

Liu, P., Huang, J., Wang, W., Xu, C., 2011. Effects of transverse rumble strips on safety of pedestrian
an
crosswalks on rural roads in China. Accident Analysis and Prevention 43, 1947-1954.

Machin, M. A., Sankey, K. S., 2008. Relationships between young drivers’ personality characteristics,
risk perceptions, and driving behavior. Accident Analysis and Prevention 40, 541-547
M
McKenna, F. P., Horswill, M. S., 2006. Risk taking from the participants’ perspective: the case of
driving and accident risk. Health Psychology 25, 163–170.
ed

National Development and Reform Commission, 2008. The outline of the plan for the reform and
development of the Pearl River Delta 2008-2020.

Organisation for Economic Co-operation and Development, 2006. Young drivers: The road to safety.
pt

Retrieved from http://www.internationaltransportforum.org/jtrc/safety/YDpolicyBrief.pdf.

Paaver, M., Eensoo, D., Kaasik, K., Vaht, M., Mäestu, J., Harro, J., 2012. Preventing risky driving: A
ce

novel and efficient brief intervention focusing on acknowledgement of personal risk factors. Accident
Analysis and Prevention 50, 430–437.

Ponce, J. de C., Munoz, D. R., Andreuccetti, G., Carvalho, D. G. de C., Leyton, V., 2011.
Ac

Alcohol-related traffic accidents with fatal outcomes in the city of Sao Paulo. Accident Analysis and
Prevention 43, 782-787.

Poulter D. R., Chapman, P., Bibby, P. A., Clarke, D. D., Crundall, D., 2008. Relationships between
young drivers' personality characteristics, risk perceptions, and driving behavior. Accident Analysis and
Prevention 40, 2058–2064.

Routley, V., Ozanne-Smith, J., Li, D., Yu, M., Wang, J., Zhang, J., Tong, Z., Wu, M., Wang, P., Qin, Y.,
2008. China belting up or down? Seat belt wearing trends in Nanjing and Zhoushan. Accident Analysis
and Prevention 40, 1850-1858.

Shinar, D., Compton, R., 2004. Aggressive driving: an observational study of driver, vehicle, and
situational variables. Accident Analysis and Prevention 36, 429-437.

Page 23 of 24
Smart, R. G., Mann, R. E., 2002. Deaths and injuries from road rage: Cases in Canadian newspapers.
Canadian Medical Association Journal 167 (7), 761–762.

Tay, R., 2005. The effectiveness of enforcement and publicity campaigns on serious crashes involving
young male drivers: Are drink driving and speeding similar? Accident Analysis and Prevention 37,
922-929.

Traffic Management Bureau, Ministry of Public Security. 2012 Annual statistical report on road traffic
accidents.

t
Tran, N. T., Bachani, A. M., Pham, V. C., Lunnen, J. C., Jo, Y. J., Passmore, J., Nguyen, P. N., Hyder, A.

ip
A., 2012. Drinking and driving in Vietnam: Public knowledge, attitudes, and practices. Traffic Injury
Prevention 13, 37-43.

cr
Tranter, P., Warn, J., 2008. Relationships between interest in motor racing and driver attitudes and
behaviour amongst mature drivers: An Australian case study. Accident Analysis and Prevention 40,
1683–1689.

us
WHO. 2004. World report on road traffic injury prevention.

Wickens, C. M., Mann, R. E., Stoduto, G., Ialomiteanu, A., Smart, R. G., 2011. Age group differences
an
in self-reported aggressive driving perpetration and victimization. Transportation Research Part F 14,
400-412.

Wickens, C. M., Toplak, M. E., Wiesenthal, D. L., 2008. Cognitive failures as predictors of driving
M
errors, lapses, and violations. Accident Analysis and Prevention 40, 1223–1233.

Yannis, G., Papadimitriou, E., Antoniou, C., 2007. Multilevel modelling for the regional effect of
enforcement on road accidents. Accident Analysis and Prevention 39, 818-825.
ed

Zakletskaia, L. I., Mundt, M. P., Balousek, S. L., Wilson, E. L., Fleming, M. F., 2009. Alcohol-impaired
driving behavior and sensation-seeking disposition in a college population receiving routine care at
campus health services centers. Accident Analysis and Prevention 41 (3), 380–386.
pt

Zhang, W., Huang, Y. H., Roetting, M., Wang, Y., Wei, H., 2006. Driver's views and behaviors about
safety in China - What do they NOT know about driving? Accident Analysis and Prevention 38, 22-27.
ce

Zhang, G., Yau, K. K. W., Chen, G., 2013. Risk factors associated with traffic violations and accident
severity in China. Accident Analysis and Prevention 59, 18-25.
Ac

Zhuang, X., Wu, C., 2011. Pedestrians' crossing behaviors and safety at unmarked roadway in China.
Accident Analysis and Prevention 43, 1927-1936.

Page 24 of 24