The effects of traffic wardens on the red-light infringement behavior of vulnerable road users

Transportation Research Part F 37 (2016) 52–63

Contents lists available at ScienceDirect

Transportation Research Part F journal homepage: www.elsevier.com/locate/trf

The effects of traffic wardens on the red-light infringement behavior of vulnerable road users Xiaobao Yang a,b,⇑, Mohamed Abdel-Aty b, Mei Huan a, Bin Jia a, Yichuan Peng b a b

MOE Key Laboratory for Urban Transportation Complex Systems Theory and Technology, Beijing Jiaotong University, Beijing 100044, China Department of Civil, Environmental & Construction Engineering, University of Central Florida, Orlando, FL 32816, United States

a r t i c l e

i n f o

Article history: Received 28 November 2014 Received in revised form 4 August 2015 Accepted 13 December 2015

Keywords: Traffic warden Vulnerable road users Red-light infringement Road safety

a b s t r a c t In some Chinese cities, traffic wardens are employed to maintain traffic order at the intersections with a high mixed traffic flow in peak hours. The main work of a traffic warden is to advise pedestrians and riders to wait at the appropriate area during red light periods. In many other countries, there are also traffic wardens at the areas with dense crowds or vehicles, such as at some large parking lots near parks and shopping malls. This paper investigated the effects of traffic wardens on the crossing behavior of pedestrians, cyclists and electric bike riders at signalized intersections. A total of 795 samples with traffic wardens and 773 samples without traffic wardens at intersections in Beijing, China were observed. Logistic regression and multivariate analysis of variance were used to test the effect of traffic wardens on the red-light crossing behavior. The results indicated that the presence of traffic wardens would significantly reduce by 21% red-light infringement behavior of vulnerable road users. However, the effects of traffic wardens were different among different groups. The effect of traffic wardens on reducing the violation rate for pedestrians was smaller compared to cyclists and electric bike riders (8% vs. 23% and 27%, respectively). The effect of traffic wardens was significant for the individuals approaching straight ahead the intersection, while it was not significant for the individuals approaching from the left and right sides of the observed direction. The possible reasons for the different effects and some practical countermeasures to reduce red light infringement were discussed. This study provides insights into the effects of traffic wardens on enhancing the safety of vulnerable road users under mixed traffic conditions. It is useful for the administrators to evaluate the supervision performance of traffic wardens and make informed decisions to employ traffic wardens at locations with dense crowds or vehicles. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction The current road networks are planned and designed largely from a car-user perspective, which leads to the lack of adequate facilities for slow moving traffic (Li, 2013; Tiwari, Bangdiwala, Saraswat, & Gaurav, 2007). However, pedestrians and cyclists are regarded as vulnerable road users as they are hardly protected in traffic collisions (ETSC, 1999). Vulnerable road users involved injuries and fatalities are overrepresented in traffic accidents. Worldwide, pedestrians and two-wheeled

⇑ Corresponding author at: MOE Key Laboratory for Urban Transportation Complex Systems Theory and Technology, Beijing Jiaotong University, Beijing 100044, China. E-mail address: [email protected] (X. Yang). http://dx.doi.org/10.1016/j.trf.2015.12.009 1369-8478/Ó 2015 Elsevier Ltd. All rights reserved.

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

53

riders contribute to 46% of the total road traffic deaths (WHO, 2009). In 2010, 16,281 pedestrians and 10,653 riders of nonmotor vehicles were recorded as killed in road accidents in China, representing 25.0% and 16.3% of road traffic fatalities, respectively (CRTASR, 2011). In Delhi, India, pedestrians and two-wheeled riders accounted for 70–85% of the total traffic deaths in the period 1990–2004 (Tiwari et al., 2007). In the developed countries, the proportion of road accidents involving pedestrians and cyclists are also high, though they have a relatively low rate of traffic fatalities. For example, in the Netherlands, there were a total of 19,100 persons seriously injured in traffic accidents in 2010, among which around 50% were caused by bicycle accidents (IRTAD, 2013). These data suggest that not enough effort is being done to meet the needs of vulnerable road users. Fortunately, the safety of vulnerable road groups has been more of a concern to many researchers. In 2012, a special issue about vulnerable road users was published in Accident Analysis & Prevention (Shinar, 2012). However, many existing literature were conducted on the unsafe behavior, safety perception, safety evaluation, accidents and injuries. Relatively, only a few studies have been focused on how to improve the safety of vulnerable road users. As to the countermeasures of improving pedestrian safety, Retting, Ferguson, and McCartt (2003) summarized the engineering countermeasures and classified them into three groups: separation of pedestrians from vehicles, speed control of motor vehicle, and increase of pedestrians’ visibility and conspicuity. A useful review on pedestrian safety education can be found in Duperrex, Bunn, and Roberts (2002). In addition, Federal Highway Administration (FHWA) published several reports about special measures to improve pedestrian safety, such as traffic calming and countdown pedestrian signals (Ewing, 1999; Singer & Lerner, 2005). Recently, Zegeer and Bushell (2012) proposed some potential strategies which included providing pedestrian-friendly geometric guidelines, promoting the use of enforcement, development and implementation of effective traffic control, safety education programs, safer bus stop and school routes, etc. Lipovac, Vujanic, Maric, and Nesic (2013a, 2013b) indicated that a countdown display would significantly reduce the total number of offenders. However, this reduction is not the same at various categories of pedestrians. Fredriksson and Rosén (2012) indicated that integrated systems of passive and active pedestrian countermeasures such as bonnets/airbags and autonomous braking can significantly reduce the likelihood of head injury compared to either of the two systems alone. Wei, Becic, Edwards, Graving, and Manser (2014) proposed several technology-based and infrastructure-based solutions to reduce the collisions between pedestrians and transit buses. In addition, waiting time was identified as a significant variable in pedestrians’ red-light crossing behavior (Brosseau, Zangenehpour, Saunier, & Miranda-Moreno, 2013; Guo, Gao, Yang, & Jiang, 2011; Tiwari et al., 2007). Decreasing pedestrians’ waiting times is likely to diminish dangerous crossing behaviors and the risk of car-pedestrian collision at signalized intersections. As to the countermeasures of improving cycling safety, firstly, most measures for pedestrians can be correspondingly applied to enhance cyclists’ safety, for example, educational programs, enforcement of speed limits, and some infrastructure measures such as traffic calming and the technique of space-sharing (ETSC, 2012; Wegman, Zhang, & Dijkstra, 2012). Secondly, some special measures for cyclists were put forward. For instance, some researchers focused on the effects of helmet use on reducing cycling injury (Dellinger & Kresnow, 2010; Fyhri, Bjørnskau, & Backer-Grøndahl, 2012). Wood et al. (2012) found that the use of reflective clothing can improve the conspicuity of bicyclists, while the presence of a bicycle light did not improve the conspicuity at night. Zhang and Wu (2013) indicated that sunshields installed at intersections can decrease by 5.5% and 4.8% the violation rates of two-wheeled riders on sunny and cloudy days, respectively. Wegman et al. (2012) suggested that we needed a paradigm shift to a sustainable safe traffic system which had five central principles: functionality, homogeneity, predictability, forgivingness and state awareness. In this study, we explore another countermeasure-which is traffic wardens, and test their effects on the crossing behavior of pedestrians and bike riders at urban intersections in China. A similar study was conducted by Rosenbloom, Haviv, Peleg, and Nemrodov (2008). They used a questionnaire survey to evaluate the effectiveness of the crossing guard (i.e. traffic warden) program in Israeli elementary schools. The results indicated that the crossing guard program had a significant influence in some domains and within certain age groups. Owing to the weakness of safety awareness and enforcement, red-light infringement behavior of pedestrians and bike riders is rather prevalent and represents a substantial safety problem at Chinese urban intersections (Wu, Yao, & Zhang, 2012; Yang, Deng, Wang, Li, & Wang, 2006). In addition, the intersections on the arterial streets are key nodes in urban traffic system and influence the safety and efficiency of the whole urban network. Therefore, traffic wardens are often employed to maintain traffic order at these crucial intersections in peak hours on weekdays in China. They usually stand near the intersection stop line for motor and non-motor vehicles, dressed in reflective clothing with a red flag in one hand (see Fig. 1). Although this measure requires manpower, it should effectively reduce the likelihood of red-light infringement behavior and relieve traffic congestion at urban intersections, which further improves the safety and efficiency of the whole urban network. The main work of a traffic warden is to advise pedestrians and bike riders approaching the intersection during red light periods to wait at the appropriate area until the traffic light turns green. Apart from supervising the bike riders, traffic wardens sometimes blow the whistle to prevent pedestrians from infringement on the traffic signal. In addition, the presence of traffic wardens may make psychological pressures on the infringement behavior no matter where he/she waits. However, unlike traffic police, traffic wardens are not qualified to issue traffic tickets. Consequently, some people would not comply with the supervision of traffic wardens and cross against the traffic light. Therefore, it is necessary to evaluate the supervision performance of traffic wardens. It can be done by comparing the difference in red-light infringement behavior under different situations with and without wardens.

54

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

Fig. 1. Photograph of traffic warden at an intersection in Beijing, China.

In many other countries, there are also traffic wardens at some areas with dense crowds or vehicles. A traffic warden usually makes the gesture or holds a flag or glow stick to guide the flow of traffic. For example, the crossing guard program was implemented on the roads near schools in Israel (Rosenbloom et al., 2008). In addition, traffic wardens are dedicated to guiding the pedestrian flow during some large-scale sport and entertainment events. Traffic wardens are also employed to maintain traffic order of motor vehicles at some large parking lots near parks and shopping malls. Till now, little literature has been conducted on traffic wardens in China or other countries. We do not understand how traffic wardens influence road safety and traffic congestion, what percentage of people are willing to comply with the supervision of traffic wardens, when we need the supervision of traffic wardens, and how many people should be employed under various conditions. Answers to these questions should help us to enhance road safety and relieve traffic congestion at the areas with dense crowds and vehicles. The main objective of this study is to evaluate the supervision performance of traffic wardens on reducing the red-light infringement behavior at signalized intersections. According to the field observation, we predicted that (a) the presence of traffic wardens would significantly reduce the rate of red-light infringement behavior of vulnerable road users, but (b) the effects of traffic wardens would be different among traffic mode groups (pedestrians, cyclists and e-bike riders) and (c) among approaching direction groups (approaching straight ahead the intersection and approaching the intersection from the left and right sides of the observed direction). The results of this research might help to provide new information to enhance the safety and efficiency of the intersections with a high mixed traffic flow, which is a major issue in some developing countries, such as India and China. In addition, this study is useful for the administrators to evaluate the supervision performance of traffic wardens and make decisions to employ traffic wardens at the locations with dense crowds or vehicles in many other countries. 2. Method 2.1. Test sites A cross-sectional observational study was conducted at four signalized intersections in Beijing. As the aim of the study is to compare the red-light crossing behavior under the situations with and without traffic wardens, three criteria were applied to choose the intersections. Firstly, there should be traffic wardens in the chosen sites in peak hours and no traffic wardens in non-peak hours. Secondly, the chosen sites are typical representations of signalized intersections in urban areas. Thirdly, there should be a considerably high volume of motor and non-motor traffic. Before the final list of intersections was selected, fifteen intersections were observed and tested. First, the intersections where there were always traffic wardens or no traffic wardens during the whole daytime (from 7:00 a.m. to 7:00 p.m.) were excluded. Four typical intersections in Haidian District, Beijing were finally selected as the observational sites after the pilot. Table 1 summarizes the geometrical design and traffic signal characteristics of the four observed sites.1 These chosen intersections are four-arm and have separate non-motor lanes. There are traffic wardens to assist in regulating traffic order in the intersections chosen in peak hours (7:00–10:00 a.m. and 4:00–7:00 p.m.) on weekdays. There is one traffic warden on each arm of the chosen intersections. All the sites are at urban arterial streets where there are a high volume of traffic and a high number of conflicts among pedestrians, riders and drivers. In addition, there are no refuge islands which can be used as a safe refuge by pedestrians and riders at these sites.

1 This is a special red light status in China indicating that the light will change from green to red status with a yellow light for vehicles and a flashing green light for pedestrians during the following 4 s after the green light. Chinese traffic rules stipulate that vehicles and pedestrians are not allowed to enter the intersection during the yellow light or the flashing green light.

55

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63 Table 1 Summary of the four intersections observed.

a b c d

Intersection site

Site Z–X Zaojunmiao Rd.– Xueyuan South Rd.

Site W–Z Weigongcun Rd.– Zhongguancun South Rd.

Site X–Z Xueyuan Rd.– Zhichun Rd.

Site Z–N Zhongguancun Rd.– Northern Third Ring Rd.

Observed direction Motor vehicle lanes Number of phases Signal phase (s) in a cycle time of traffic lights for motor vehicles

North–South 6(N–S), 4(W–E) 3 36 + 4 + 2a 24 + 4 + 2 54 + 4 + 2

West–East 8(N–S), 4(W–E) 3 48 + 4 + 2 35 + 4 + 2 40 + 4 + 2

Signal phase (s) for pedestrians and non-motor vehicles in the observed direction

36b 4c 92d

40 4 97

South–North 5(N–S), 4(W–E) 4 38 + 4 + 2 50 + 4 + 2 18 + 4 + 2 30 + 4 + 2 38 4 118

North–South 8(N–S), 6(W–E) 4 65 + 4 + 2 40 + 4 + 2 55 + 4 + 2 30 + 4 + 2 65 4 145

36 is the green light time, 4 is the yellow light time, and 2 is the all-red or clearance time in the first signal phase. It is the green light time for pedestrians and non-motor vehicles in the observed direction. It is the yellow light time for non-motor vehicles or the flashing green light time for pedestrians in the observed direction. It is the red light time for pedestrians and non-motor vehicles in the observed direction.

2.2. Data collection Data of crossing behavior for pedestrians and riders during red light periods including both red light and yellow or flashing green light phases were collected at these intersections in good weather in April 2014. Fig. 1 illustrates the Z–X intersection (Zaojunmiao Rd.–Xueyuan South Rd., observed direction: north–south). Fig. 2 provides the sketch map of the observed site at Z–X intersection. All the pedestrians, cyclists and e-bikers who approached straight ahead the observed site (see ‘‘A” in Fig. 2, the blue lines represent cyclists’ and e-bikers’ moving trajectories, and the green lines represent pedestrians’ moving trajectories), who approached the site from the left and right sides of the observed direction (see ‘‘L” and ‘‘R” in Fig. 2) were included in this study.2 Traffic wardens were employed at 7:00–10:00 a.m. and 4:00–7:00 p.m. at each site chosen. They stood near the stop line for non-motor vehicles to maintain the order of traffic (see ‘‘SL” in Fig. 2, the white horizontal line behind the zebra crossings represents the stop line for motor and non-motor vehicles). The cameras were hidden behind the stop line so that it would not be visible by the road users. In order to ensure the comparability between two situations with and without traffic wardens, the data collection was conducted at two time periods (9:30–10:30 a.m. and 3:30–4:30 p.m.) on weekdays. There were half an hour of traffic flow with traffic wardens and another half an hour of continuous-time traffic flow without traffic wardens at each time period. This experimental design can ensure the same road geometry and weather conditions, and the similar traffic related characteristics under the situations with and without traffic wardens. 2.3. Videotape coding All road users who entered the observed sites were recorded on video, but only the vulnerable road users who approached the sites during red light phases including red light and yellow or flashing green light phases, and traveled through the intersection were coded. All the pedestrians, cyclists and e-bike riders who approached the site from any directions (‘‘A”, ‘‘L” and ‘‘R”) were included in this analysis. Videotape data was coded in the laboratory by Master students who had been professionally trained. Two data files in a pre-designed format were made for each site. One was for pedestrians and riders, and the other was for motor vehicles which may conflict with the vulnerable road users. The data file for pedestrians and riders was used to record their personal characteristics and crossing behavior information. The variables included traffic mode, gender, estimated age group, the arrival time, the departure time, the state of the traffic signal at each time, and the approaching direction. In addition, the data file for motor vehicles was used to record environmental factors related to motor traffic, including signal timing, motor vehicle volume during the red light cycle for pedestrians and riders, and the presence of traffic wardens. Finally, two data files were matched according to the arrival time and the corresponding signal cycle of each individual.

2 As shown in Fig. 2, pedestrians and riders on the left side may approach the observed site when traffic light in the horizontal direction is green while traffic light in the vertical/observed direction is red. Some cyclists and electric bike riders approach the observed site from the right side. They make a retrograde motion. Riders’ retrograde motion is an infringement behavior, but it is relatively prevalent in China, partly because of the inadequate road design. That also may be partly because some people want to cut down the distance and the time to the destination.

56

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

Fig. 2. Sketch map of the observed site at Z–X intersection.

2.4. Statistical analysis The outcome measure was red-light infringement (yes/no) in order to examine the difference in the crossing behavior under different situations with and without traffic wardens. The crossing behavior was defined as red-light infringement if a pedestrian or rider began to enter the motor lane for crossing traffic flow during red light periods. Logistic regression analysis was used to analyze the likelihood of red-light infringement and associated risk factors. In addition, the multivariate analysis of covariance was used to further analyze the effect of traffic wardens on the red-light infringement behavior. The following variables, as shown in Table 2, are adopted to construct the models. 3. Results 3.1. Descriptive statistics A total of 1568 valid observations (795 observations with wardens and 773 observations without wardens) approaching the intersections during red light periods were observed. As shown in Table 3, the descriptive statistics for the observations with wardens and without wardens were separately cross-tabulated with traffic light compliance (Yes/No). Firstly, a majority of observations were males within the gender group (563/795 = 71% at intersections with wardens, 561/773 = 73% at intersections without wardens), young and middle-aged in the age group (87% and 88% at intersections with and without wardens, respectively), and straight ahead in the approaching directions (56% and 59%, respectively). The proportion of e-bike riders was slightly higher than that of pedestrians or cyclists at intersections. Secondly, the total numbers of observations at intersections with wardens were not significantly different from those without wardens (795 vs. 773, p > .10, non-parametric binomial test). Moreover, there was no statistically significant difference in the distribution of the gender group between two situations (v2(1) = 0.512, p > .10), as well as the distributions of the age group (v2(2) = 1.157, p > .10) and the approaching direction group (v2(2) = 2.988, p > .10). However, the distributions of the traffic mode group between two situations were slightly different (v2(2) = 8.654, p<.05). This is because the proportions of travel mode may change a little between peak hours and non-peak hours. Table 3 also presents the proportions of red-light infringement (i.e. non-compliances with the traffic light). A Chi-square analysis was used to test the differences in the violation rates among different groups. First, the overall rate of red-light infringement with traffic wardens was significantly lower than the violation rate without traffic wardens (35% vs. 56%, v2(1) = 66.696, p < 0.001). In general, the violation rate in each categorical variable with wardens was lower than that without wardens. However, the violation rates of the individuals approaching from the two sides between two situations (with and without wardens) had no significant difference. (the left side: 40% vs. 48%, v2(1) = 1.816, p > .10; the right side: 52% vs. 59%, v2(1) = 1.919, p > .10). Second, e-bike riders had a higher rate of red-light infringement than cyclists and pedestrians only under the situation without wardens when other factors were not controlled for (62% vs. 54% and 48%, v2(2)

57

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63 Table 2 Definitions of variables in the models. Variables

Descriptions

Dependent variable Crossing behavior

Compliance with traffic light: 0, red-light infringement: 1

Independent variables Traffic warden Traffic mode Approaching direction

Without warden:0, with warden:1 Electric bike: 0, pedestrian: 1, regular bike: 2 Straight ahead:0, the left side: 1, the right side: 2

Other factors Gender Age group Cross traffic volume

Female: 0, male: 1 Estimated age group: elderly (>50):0, middle-aged (30–50):1, young (<30):2 Average crossing motor vehicle volume per lane per min during the red light phase when the individual arrives

Table 3 Compliances with traffic light in each descriptive category with and without traffic wardens. Management type

With traffic wardens

Compliance

Yes

No

Traffic mode E-bike riders Cyclists Pedestrians

201(66%) 156(69%) 158(60%)

105(34%) 71(31%) 104(40%) 3.956

368(65%) 147(63%)

v

306 227 262

133(38%) 100(46%) 109(52%)

214(62%) 118(54%) 99(48%) 10.770⁄⁄

347 218 208

195(35%) 85(37%) 0.289

563 232

238(42%) 104(49%)

323(58%) 108(51%) 2.744#

561 212

167(63%) 282(66%) 66(63%)

97(37%) 145(34%) 38(37%) 0.646

264 427 104

132(48%) 169(42%) 41(44%)

144(52%) 234(58%) 53(56%) 2.321

276 403 94

327(73%) 100(60%) 88(48%)

118(27%) 68(40%) 94(52%) 38.337⁄⁄⁄

445 168 182

197(43%) 71(52%) 74(41%)

259(57%) 66(48%) 106(59%) 4.109

456 137 180

515(65%)

280(35%)

795

342(44%)

431(56%)

773

v2 Approaching direction Straight ahead Left side Right side

v2 Overall ⁄

p < .05,

⁄⁄

p < .01,

⁄⁄⁄

Total

No

2

Age group Young Middle-aged Elderly

Without traffic wardens Yes

v2 Gender Male Female

Total

#

p < .001, .05 < p < .10 and otherwise p P .10.

= 10.770, p < .01), while the violation rates among three traffic modes had no significant difference when traffic wardens were present (34% vs. 31% and 40%, v2(2) = 3.956, p > .05). The individuals approaching straight ahead the site were less likely to infringe on the traffic light than those approaching from the left and right sides only under the situation with wardens (27% vs. 40% and 52%, v2(2) = 38.337, p<.001). In addition, the waiting area is considered as appropriate if a rider waits behind the intersection stop line or a pedestrian waits at the sidewalk during the red light period. Our survey data showed the individuals from both sides hardly waited at the appropriate areas (with wardens: 3%, without wardens: 4%), while a relatively high proportion of the individuals approaching straight ahead the intersection waited at the appropriate areas (with wardens: 63%, without wardens: 32%). 3.2. Logistic regression analysis As shown is Table 2, the crossing behavior (compliance with traffic light: 0 and red-light infringement: 1) was taken as the dependent variable. Traffic warden, traffic mode, approaching direction, and other category variables were used as independent variables. All these variables were included in a binary logistic regression model (see Table 4). The results showed that traffic warden was significantly associated with red-light infringement in the logistic model. The observations had 0.419 times odds of infringement at intersections with wardens compared with the intersections without wardens, when other factors were controlled for. Both traffic mode and approaching direction were found to be significant variables for predicting the red-light crossing behavior. Pedestrians and cyclists were 0.585 and 0.738 times less likely to infringe on traffic lights than e-bike riders,

58

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

Table 4 Estimated results of logistic model for the crossing behavior. Adjusted odds

Wald v2

Traffic warden With vs. without warden

0.419

0.340–0.516

66.957⁄⁄⁄

Traffic mode Pedestrian vs. e-biker Cyclist vs. e-biker

0.585 0.738

0.429–0.798 0.567–0.962

11.464⁄⁄ 5.063⁄

Approaching direction Left vs. straight Right vs. straight

1.414 2.287

1.046–1.913 1.706–3.067

5.072⁄ 30.541⁄⁄⁄

Gender Male vs. female

1.045

0.820–1.331

0.125

Age group Young vs. elderly Middle-aged vs. elderly

0.848 0.888

0.600–1.199 0.636–1.239

0.872 0.489

1.744 1.402 13.715

1.301–2.339 1.089–1.806

13.826⁄⁄⁄ 6.851⁄⁄

Cross traffic volume Low vs. high Medium vs. high Hosmer and Lemeshow test ⁄

95% C.I. for odds

p < .05,

⁄⁄

p < .01,

⁄⁄⁄

p < .001, and otherwise p P .05.

respectively, when other variables were controlled for. The observations approaching the site from the left and right sides were 1.414 and 2.287 times more likely to infringe on traffic lights than those approached straight ahead the site. In addition, cross traffic volume had a significant effect on the red-light infringement behavior of vulnerable road users. Under low and medium levels of cross traffic volume, the likelihoods of red-light infringement were 1.744 and 1.402 times compared to high level of cross traffic volume. But the result showed both age group and gender were insignificant variables for predicting the crossing behavior in the logistic model, when other variables were controlled for. Furthermore, three logistic models were constructed separately for pedestrians, cyclists and e-bike riders to explore the difference among the three groups. The results showed that traffic warden was all identified as significant variables on these three models. Pedestrians had 0.756 times odds of infringement at intersections with wardens compared with the intersections without wardens, when other factors were controlled for. The corresponding figures were 0.347 and 0.299 times for cyclists and e-bike riders, respectively. These results are supported by the descriptive statistics shown in Table 3. For pedestrians, the rate of red-light infringement at intersections with wardens was 8% less than that at intersections without wardens. For cyclists and e-bike riders, the reductions in the violation rate were 23% and 27%, respectively. 3.3. Multivariate analysis of variance In this section, multivariate analysis of covariance (MANCOVA) was used to further analyze the effect of traffic wardens on the violation behavior of vulnerable road users. The rate of red-light infringement behavior was taken as the dependent variable. Traffic warden, traffic mode and approaching direction were considered as between-subjects factors, while cross traffic volume was used as covariant. Table 5 presents the result of MANCOVA. Main effects of three independent variables including traffic warden (F(1, 1549) = 28.745, p < 0.001), traffic mode (F(2, 1549) = 9.239, p < 0.001), and approaching direction (F(2, 1549) = 8.852, p < 0.001) were significant. The interaction effect of traffic warden  approaching direction (F(2, 1549) = 6.255, p = 0.002) was significant for the rate of red-light crossing behavior. No other interaction effects were found to be significant. In addition, the covariant, namely cross traffic volume, were identified to have a significant effect on the rate of red-light infringement (F(1, 1549) = 20.007, p < 0.001). The MANCOVA was constructed separately for pedestrians, cyclists and e-bike riders to explore the difference among the traffic mode groups. The results showed that traffic warden was significant in all three separate models for pedestrians (F(1, 463) = 5.951, p = 0.015), cyclists (F(1, 438) = 8.091, p = 0.005) and e-bike riders (F(1, 646) = 14.888, p < 0.001). The MANCOVA was also constructed separately for the individuals approaching from different directions. The results showed that the effect of traffic wardens for the individuals approaching straight ahead the site was significant (F(1, 894) = 54.128, p < 0.001). However, the effects of traffic wardens for the individuals approaching the site from the left side (F(1, 298) = 3.801, p = 0.052) and the right side (F(1, 355) = 1.922, p = 0.167) were all insignificant. Furthermore, the MANCOVA was conducted with gender and age group input as independent variables as well to avoid any potential confounding effects from these factors. The results showed the effects of gender (F(1, 1533) = 0.184, p = 0.668), and age group (F(2, 1533) = 0.796, p = 0.451) on the red-light infringement behavior were insignificant.

59

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63 Table 5 Estimated results of multivariate analysis of variance for the crossing behavior. Source

Sums of squares

Degrees of freedom

Mean squares

F

p

Traffic warden Traffic mode Approaching direction Warden  mode Warden  direction Mode  direction Warden  mode  direction Cross traffic volume Error

6.520 4.191 4.015 0.458 2.837 1.614 0.854 4.538 351.340

1 2 2 2 2 4 4 1 1549

6.520 2.096 2.008 0.229 1.419 0.403 0.213 4.538 0.230

28.745 9.239 8.852 1.009 6.255 1.779 0.941 20.007

<0.001 <0.001 <0.001 0.365 0.002 0.130 0.439 <0.001

Total

388.601

1567

3.4. Crossing behavior characteristics Pedestrians’ and riders’ street-crossing behavior can be classified into three distinct manners: risk-taking, opportunistic, and law-obeying (Wu et al., 2012; Yang et al., 2006). The risk-taking ones would cross against the red light without any waiting after approaching the intersection. The opportunistic ones would initially wait at the red light but would become impatient and subsequently infringe on the red light by seeking gaps among crossing traffic. In this section, the percentages of risk-taking and opportunistic ones between two situations with and without wardens are compared and analyzed. Table 6 presents the percentages of risk-taking ones among different (traffic mode and approaching direction) groups under two situations with and without wardens. First, pedestrians had a lower likelihood of risk-taking behavior than cyclists and e-bike riders under any situation with and without wardens. Second, traffic wardens can significantly reduce the percentages of risk-taking behavior among each traffic mode group. The reductions for pedestrians, cyclists and e-bike riders were 8.0%, 6.0% and 6.3%, respectively. Third, the effects of traffic wardens on deducing the risk-taking behavior was significant for the individuals approaching straight ahead the site, but insignificant for the individuals approaching from both sides of the observed direction. Table 7 presents the percentages of opportunistic ones among each group under two situations. There was no significant difference in the percentages of opportunistic pedestrians between two situations with and without wardens. However, traffic wardens can significantly reduce the percentages of opportunistic behavior among the cyclist and e-biker groups. The reductions for cyclists and e-bike riders were 16.8% and 21.0%, respectively. In addition, there was significant difference in the percentages of opportunistic ones for the individuals approaching straight ahead the site. However, there were no significant differences in the percentages of opportunistic ones for the individuals approaching from both sides. 3.5. Effects of traffic wardens for different sites Table 8 presents the results of the significances of the difference in the red-light violation rates between two situations with and without wardens at each observed site. The results show that on the whole, traffic wardens can significantly reduce Table 6 The percentages of risk-taking ones among each group.

Pedestrians Cyclists E-bikers Straight Left side Right side ⁄

p < .05,

⁄⁄

p < .01,

⁄⁄⁄

No wardens (%)

With wardens (%)

Differences (%)

Z-test

13.0 18.8 16.4 16.4 5.8 23.3

5.0 12.8 10.1 8.1 3.6 17.0

8.0 6.0 6.3 8.3 2.2 5.3

2.982⁄⁄⁄ 1.748⁄ 2.391⁄⁄ 3.861⁄⁄⁄ 0.921 1.497

p < .001, and otherwise p P .05.

Table 7 The percentages of opportunistic ones among each group.

Pedestrians Cyclists E-bikers Straight Left side Right side ⁄

p < .05,

⁄⁄

p < .01,

⁄⁄⁄

No wardens (%)

With wardens (%)

Differences (%)

34.6 35.3 45.2 40.4 42.3 35.6

34.7 18.5 24.2 18.4 36.9 34.6

0.1 16.8 21.0 12.0 5.4 1.0

p < .001, and otherwise p P .05.

Z-test 0.027 4.065⁄⁄⁄ 5.812⁄⁄⁄ 7.452⁄⁄⁄ 0.965 0.187

60

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

Table 8 Can traffic wardens reduce significantly the red-light violation rate at each site?.

Straight Left side Right side Over all

Site 1

Site 2

Site 3

Site 4

Y N N Y

Y N N Y

Y N N Y

Y N N Y

Y-yes, N-no.

the violation rate for each observed site. Based on the samples at each site, we obtained the same results that the effect of traffic wardens on reducing the red-light crossing behavior was significant for the individuals approaching straight ahead, but insignificant for the individuals approaching from both sides. 4. Discussion Owing to the lack of safety awareness and enforcement, red-light crossing behavior of pedestrians and bike riders at signalized intersections is a common phenomenon. Our survey data showed there were high rates of red-light infringement behavior at intersections without wardens (pedestrians: 48%; two-wheelers: 59%), which is similar as another empirical result by Wu et al. (2012) in Beijing (the mean violation rate of two-wheelers in their observations is 56%). On the other hand, the intersections on the arterial streets are key nodes in urban traffic system and influence the safety and efficiency of the whole urban network. In this case, traffic wardens are employed to maintain traffic order at arterial intersections in peak hours on weekdays in China. The objective of this study was to explore the effects of traffic wardens on decreasing the violation rates of vulnerable road users at signalized intersections. The data collection was conducted at twelve one-hour periods when there were half an hour with traffic wardens and another half an hour without traffic wardens at each one-hour period. This experimental design can ensure the same road geometry and weather conditions, and the similar traffic related characteristics between two cases with and without traffic wardens. In addition, the number of the samples had no significant difference between two situations with and without wardens, as well as the distributions of the gender group, the age group, and the approaching direction group. The results of this study identified that traffic wardens can significantly reduce by 21% the rates of the red-light infringement behavior. Although this measure requires manpower, it can effectively reduce the likelihood of redlight infringement behavior and relieve traffic congestion at urban intersections with a high mixed traffic flow, which is a major issue in some developing countries, such as India and China. The logistic regression analysis and multivariate analysis of covariance were applied to test the difference in the red-light violation rates between two cases with and without wardens. Traffic warden, traffic type and approaching direction were found to have significant effects on the red-light crossing behavior, when other factors were controlled for. Results of the logistic regression showed the odds of infringement at intersections with wardens would be reduced to 0.419 times compared with the situation without wardens, which proved a significantly positive performance of traffic wardens in supervising the red-light infringement behavior of vulnerable road users. Moreover, the significant effects of traffic wardens on reducing the violation rates were identified separately for pedestrians, cyclists and e-bike riders. However, the reduction in the red-light violation rates was significantly different among three travel mode groups. Pedestrians were 0.756 times less likely to infringe on the red light with wardens than those without wardens. The corresponding figures were 0.347 and 0.299 times for cyclists and e-bike riders, respectively. Accordingly, it was revealed that the effect of traffic wardens on reducing the violation rates for pedestrians was smaller compared to cyclists and electric bike riders. The different decreasing rates of red-light infringement with wardens mainly resulted from the differences in the initial violation rates without wardens. Under the situation without wardens, e-bike riders were more likely to infringe on traffic lights than cyclists and pedestrians (62% vs. 54% and 48%, respectively), while there were no significant difference in the final violation rates among the three groups when traffic wardens were present (34% vs. 31% and 40%, respectively). Therefore, the findings indicate that, under the supervision of traffic wardens, the final violation rates had no significant difference among the three groups, though the reductions in the violation rates were different. Furthermore, the result of the MANCOVA showed that the interaction effect of traffic warden  approaching direction was statistically significant. This significant interaction effect indicated that the effect of traffic wardens on the red-light infringement behavior was dependent on the approaching direction. In other words, the supervision performances of traffic wardens were significantly different for the individuals approaching the observed site from different directions. Further data analysis separately for pedestrians, cyclists and e-bike riders revealed that the effect of traffic wardens for the individuals approaching straight ahead the site was significant. However, traffic wardens had no significant effects on reducing the violation rates for the individuals approaching from the left and the right sides. The main reason may be associated with the difference in the waiting positions of the individuals. As shown in Fig. 2, the blue and green lines represent their moving trajectories. The individuals approaching straight ahead need to pass the intersection stop line. A traffic warden would wave a flag to advise them to wait behind the stop line. Also, their crossing

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

61

behaviors are impacted by other people waiting behind the stop line. People are more likely to waiting for the green light when traffic wardens are present compared with the situation without wardens. Based on the theory of social control, those surrounded by other people waiting for the green light feel more committed to social order and tend to comply with the law (Rosenbloom, 2009; Wu et al., 2012). Consequently, the individuals approaching straight ahead are less likely to infringe on traffic lights at intersections with wardens compared with the situation without wardens. However, for the individuals approaching from both sides, their whole moving trajectories are in front of the intersection stop line for non-motor and motor vehicles. In this case, their crossing behaviors are not impacted by other people waiting behind the stop line. It is also difficult for traffic wardens to supervise their crossing behavior. Therefore, the violation rates of the individuals approaching from both sides had no significant difference between two situations with and without wardens. In addition, the individuals from both sides hardly waited at the appropriate waiting area (with wardens: 3%, without wardens: 4%). The vast majority of them either cross the intersection without any waiting or waited at the edge of the crossing motor lane or in the crossing non-motor lane. Apart from the nonstop behavior, it is also unsafe to wait at the inappropriate area where they have potential conflicts with crossing traffic, though they may not cross against the red light. Therefore, safety education and enforcement programs for vulnerable road users should focus on the compliances with both the traffic light and the appropriate waiting area. For example, making an obvious mark of the appropriate waiting areas on the surface should be considered. Or the inappropriate waiting areas should be marked to alert users. The effects of other factors including gender, estimated age group and cross traffic volume were also examined. First, the descriptive statistic (see in Table 3) showed that men had slightly higher rates of red light infringement than women only under the situation without wardens. Similar results of the observation study in Beijing were reported by Guo et al. (2011), and Wu et al. (2012). Second, different age group had no significant difference on the violation rate. One of the reasons may be some differences in the subjective judgment by different age groups (Zhang & Wu, 2013). Another possible reason may be that there were several high schools and campuses in our surveyed area. A number of the young in the observations were students who had high safety consciousness (Guo et al., 2011). In addition, results of logistic regression and MANCOVA indicated crossing traffic volume had a significant effect on the rate of red-light infringement when other factors were controlled for. As crossing traffic flow increases, the likelihood of red-light infringement decreases. Similar results were obtained in the crossing behaviors of pedestrians (Guo et al., 2011; Hamed, 2001; Rosenbloom, 2009; Yagil, 2000) and bike riders (Wu et al., 2012; Yang, Huan, Abdel-Aty, Peng, & Gao, 2015; Zhang & Wu, 2013). The analysis result of the percentage of risk-takers showed that pedestrians were less likely to make a risk-taking behavior than cyclists and e-bike riders under both situations with and without traffic wardens. This is because of the difference of crossing speed among different traffic modes. Riding a regular or electric bike can cross the intersection with higher speed and less time compared to walking across the intersection. Therefore, cyclists and e-bike riders had higher likelihoods of risktaking behavior than pedestrians. In addition, traffic wardens can significantly reduce by 8.0% pedestrians’ risk-taking behavior. The corresponding figures for cyclists and e-bike riders were 6.0% and 6.3%, respectively. The results indicated that the effect of traffic wardens on reducing the percentage of risk-taking behavior for pedestrians is larger than cyclists and electric bike riders. The effects of traffic wardens on decreasing opportunistic behavior were not the same for the different traffic mode groups. Cyclists and e-bike riders had lower percentages of opportunistic ones at intersections with wardens compared with the intersections without wardens. However, the difference in percentage of the opportunistic pedestrians under the two situations with and without wardens was not significant. This is because pedestrians’ waiting position is away from traffic wardens. They would terminate their waiting behaviors and cross against the red light if an enough gap between successive cars is present. The results also indicated that the effects of traffic wardens on decreasing risk-taking and opportunistic behaviors were not the same for different approaching direction groups. This is caused by the difference in the waiting position. The individuals approaching from both sides mostly would wait in front of the stop line. It is hard for traffic wardens to supervise their crossing behavior because of the long distance. Therefore, traffic wardens had no significant effects on their risktaking and opportunistic behaviors. 5. Conclusions This paper investigated the effective effects of traffic wardens on decreasing the rate of red-light infringement behavior of pedestrians, cyclists and e-bike riders. A total of 1568 observations approaching the intersections during red light periods were observed. The following several conclusions were made: (1) The results proved that the effectiveness of traffic wardens on reducing the red-light infringement rates of vulnerable road users. Although this measure requires manpower, it is an appropriate measure to improve the safety and efficiency of key intersections with high mixed traffic flow, which is a major issue in some developing countries, such as India and China. (2) The effects of traffic wardens were not the same among three traffic mode groups. The reduction in the violation rate for pedestrians was smaller than cyclists and electric bike riders. However, the reduction in the risk-taking behavior for pedestrians was larger than cyclists and electric bike riders.

62

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

(3) There was a significant interaction effect between traffic warden and approaching direction. The result indicated the supervision performances of traffic wardens were significantly different when the individuals approached the observed site from different directions. The individuals approaching straight ahead the site had a significant reduction in the red-light infringement behavior, while the individuals approaching from two sides had no significant difference in the infringement behavior between two situations with and without wardens. (4) The individuals from both sides hardly waited at the appropriate waiting area. Safety education and enforcement programs about the street-crossing behavior should focus on the compliance with both the traffic light and the appropriate waiting area. Although this study provides insights into the effects of traffic wardens on enhancing the safety of vulnerable road users under mixed traffic conditions, there are several limitations which should be done in future work. First, we only investigated the effects of traffic wardens on the behavior of vulnerable road users in this paper. Drivers may consider that most pedestrians and bike riders would comply with traffic lights at intersections when traffic wardens are present. Thus, their alertness may be reduced with respect to the red-light infringement behavior of vulnerable road users. The effect of traffic wardens on drivers’ psychology and driving behavior should be studied. Second, the effects of traffic wardens on the traffic safety and capacity of the whole intersection should be evaluated. Traffic simulation can be used to compare the differences in traffic safety and efficiency under various traffic and environmental conditions between two situations with and without wardens, and then make the decisions where and when traffic wardens are needed. Next, it was proven that traffic wardens can significantly reduce the violation rate for each observed site. However, the effect of traffic wardens on reducing the percentage of red-light infringement may be unequal among different intersections. The effect of the specific intersection site should be further explored in future research. Finally, at some other areas with dense crowds or vehicles, such as large parking lots near parks and shopping malls, the effect of traffic wardens on traffic safety should be also evaluated. We also need to identify when and how many traffic wardens are needed under various conditions at these areas. Acknowledgements This work was supported by the National Basic Research Program of China (Grant No. 2012CB725401), National Natural Science Foundation of China (Grant Nos. 71131001, 71222101), Fundamental Research Funds for the Central Universities of China (Grant No. 2015JBM054), and Beijing Postdoctoral Research Foundation of China (Grant No. 2015zz-96). References Brosseau, M., Zangenehpour, S., Saunier, N., & Miranda-Moreno, L. (2013). The impact of waiting time and other factors on dangerous pedestrian crossings and violations at signalized intersections: A case study in Montreal. Transportation Research Part F, 21(6), 159–172. CRTASR, 2011. China road accidents statistics report in 2010. Traffic administration Bureau of China, Beijing, China. Dellinger, A. M., & Kresnow, M. (2010). Bicycle helmet use among children in the United States: The effects of legislation, personal and household factors. Journal of Safety Research, 41(4), 375–380. Duperrex, O., Bunn, F., & Roberts, I. (2002). Safety education of pedestrians for injury prevention: A systematic review of randomized controlled trials. British Medical Journal, 324, 1129–1133. ETSC (1999). Safety of pedestrians and cyclists in urban areas. Brussels, Belgium: European Transport Safety Council. ETSC (2012). Review of cycling safety policies in the European Union. Brussels, Belgium: European Transport Safety Council. Ewing, R. H. (1999). Traffic calming: State of the practice. Washington, DC, USA: Federal Highway Administration. Fredriksson, R., & Rosén, E. (2012). Integrated pedestrian countermeasures – Potential of head injury reduction combining passive and active countermeasures. Safety Science, 50(3), 400–407. Fyhri, A., Bjørnskau, T., & Backer-Grøndahl, A. (2012). Bicycle helmets – A case of risk compensation? Transportation Research Part F, 15(5), 612–624. Guo, H. W., Gao, Z. Y., Yang, X. B., & Jiang, X. B. (2011). Modeling pedestrian violation behavior at signalized crosswalks in China: A hazards-based duration approach. Traffic Injury Prevention, 12(1), 96–103. Hamed, M. M. (2001). Analysis of pedestrians’ behavior at pedestrian crossings. Safety Science, 38(1), 63–82. IRTAD (2013). Road safety annual report 2013. Paris, France: International Traffic Safety Data and Analysis Group. Li, B. B. (2013). A model of pedestrians’ intended waiting times for street crossings at signalized intersections. Transportation Research Part B, 51(1), 17–28. Lipovac, K., Vujanic, M., Maric, B., & Nesic, M. (2013a). Pedestrian behavior at signalized pedestrian crossings. Journal of Transportation Engineering, 139(2), 165–172. Lipovac, K., Vujanic, M., Maric, B., & Nesic, M. (2013b). The influence of a pedestrian countdown display on pedestrian behavior at signalized pedestrian crossings. Transportation Research Part F, 20, 121–134. Retting, R. A., Ferguson, S. A., & McCartt, A. T. (2003). A review of evidence-based traffic engineering measures designed to reduce pedestrian-motor vehicle crashes. American Journal of Public Health, 93(9), 1456–1463. Rosenbloom, T. (2009). Crossing at a red light: Behavior of individuals and groups. Transportation Research Part F, 12(5), 389–394. Rosenbloom, T., Haviv, M., Peleg, A., & Nemrodov, D. (2008). The effectiveness of road-safety crossing guards: Knowledge and behavioral intentions. Safety Science, 46(10), 1450–1458. Shinar, D. (2012). Safety and mobility of vulnerable road users: Pedestrians, bicyclists and motorcyclists. Accident Analysis and Prevention, 44(1), 1–2. Singer, J. P., & Lerner, N. D. (2005). Countdown pedestrian signals: A comparison of alternative pedestrian change interval displays. Washington, DC, USA: Federal Highway Administration. Tiwari, G., Bangdiwala, S., Saraswat, A., & Gaurav, S. (2007). Survival analysis: Pedestrian risk exposure at signalized intersections. Transportation Research Part F, 10(2), 77–89. Wegman, F., Zhang, F., & Dijkstra, A. (2012). How to make more cycling good for road safety. Accident Analysis and Prevention, 44(1), 19–29. Wei, C., Becic, E., Edwards, C., Graving, J., & Manser, M. (2014). Task analysis of transit bus drivers’ left-turn maneuver: Potential countermeasures for the reduction of collisions with pedestrians. Safety Science, 68(10), 81–88. WHO (2009). Global status report on road safety: Time for action. Geneva, Switzerland: World Health Organization. Wood, J. M., Tyrrell, R. A., Marszalek, R., Lacherez, P., Carberry, T., & Chu, B. S. (2012). Using reflective clothing to enhance the conspicuity of bicyclists at night. Accident Analysis and Prevention, 45(3), 726–730.

X. Yang et al. / Transportation Research Part F 37 (2016) 52–63

63

Wu, C. X., Yao, L., & Zhang, K. (2012). The red-light running behavior of electric bike riders and cyclists at urban intersections in China: An observational study. Accident Analysis and Prevention, 49(11), 186–192. Yagil, D. (2000). Beliefs, motives and situational factors related to pedestrians’ self-reported behavior at signal-controlled crossings. Transportation Research Part F, 3(1), 1–13. Yang, J. G., Deng, W., Wang, J. M., Li, Q. F., & Wang, Z. A. (2006). Modeling pedestrians’ road crossing behavior in traffic system micro-simulation in China. Transportation Research Part A, 40(3), 280–290. Yang, X. B., Huan, M., Abdel-Aty, M., Peng, Y. C., & Gao, Z. Y. (2015). A hazard-based duration model for analyzing crossing behavior of cyclists and electric bike riders at signalized intersections. Accident Analysis and Prevention, 74(1), 33–41. Zegeer, C. V., & Bushell, M. (2012). Pedestrian crash trends and potential countermeasures from around the world. Accident Analysis and Prevention, 44(1), 3–11. Zhang, Y. Q., & Wu, C. X. (2013). The effects of sunshields on red light running behavior of cyclists and electric-bike riders. Accident Analysis and Prevention, 52(3), 210–218.