Accepted manuscript doi:

10.1680/jmuen.17.00005
Accepted manuscript
As a service to our authors and readers, we are putting peer-reviewed accepted manuscripts
(AM) online, in the Ahead of Print section of each journal web page, shortly after acceptance.

Disclaimer
The AM is yet to be copyedited and formatted in journal house style but can still be read and
referenced by quoting its unique reference number, the digital object identifier (DOI). Once
the AM has been typeset, an ‘uncorrected proof’ PDF will replace the ‘accepted manuscript’
PDF. These formatted articles may still be corrected by the authors. During the Production
process, errors may be discovered which could affect the content, and all legal disclaimers
that apply to the journal relate to these versions also.

Version of record
The final edited article will be published in PDF and HTML and will contain all author
corrections and is considered the version of record. Authors wishing to reference an article
published Ahead of Print should quote its DOI. When an issue becomes available, queuing
Ahead of Print articles will move to that issue’s Table of Contents. When the article is
published in a journal issue, the full reference should be cited in addition to the DOI.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Submitted: 10 February 2017
Published online in ‘accepted manuscript’ format: 31 October 2017
Manuscript title: Identifying factors affecting pedestrians' crossing decisions at intersections
in Iran
Authors: Zahra Jahandideh, Babak Mirbaha and Amir A. Rassafi
Affiliation: Civil Engineering Department, Faculty of Engineering and Technology, Imam
Khomeini International University, Qazvin, Iran
Corresponding author: Babak Mirbaha, Civil Engineering Department, Faculty of
Engineering and Technology, Imam Khomeini International University, Qazvin, Iran. Tel.:
+98 21 88034045
E-mail: [email protected]

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Abstract

Walking is the basic way of movement and the most environmentally sound mode of urban transportation.

However, pedestrians sometimes put themselves in danger by engaging in risky and unsafe behaviour.

Accordingly, in order to reduce the rate of road crashes and improve pedestrians’ safety, providing preventive

countermeasures and conducting studies on pedestrians’ risk acceptance behaviours has great importance. Thus,

the aim of this study is to identify factors affecting pedestrians’ risk taking behaviour while crossing

intersections in urban streets. An observational survey of road crossings was conducted at six (Adl, Valiasr,

Ferdosi, Shohada, Khayam and Valiasr midblock) intersections located in Qazvin, Iran to determine the crossing

details of the pedestrians in both directions. Selected crossings were in near both signalised and un-signalised

intersections. 800 samples were observed and binary logit model is applied for identifying factors affecting

pedestrians’ risk taking behaviours. The results showed that the average time to collision chosen by pedestrians

were about 6.6 seconds at signalised intersections and about 5.8 seconds at un-signalised intersections. It was

also indicated that factors including individual characteristics (e.g. gender, age, dressing type, pedestrian speed,

etc.), environmental conditions (e.g. other violating pedestrians, kerb parking, waiting time, etc.) and traffic

conditions (e.g. speed of approaching vehicles, time to collision, etc.) can significantly affect pedestrians’ risk

taking behaviours.

Keywords: Risk and probability analysis; Town and city planning; Transport management.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
1. Introduction
According to the Traffic Safety reports, a pedestrian is killed every 2 hours and one is injured every 8
minutes in traffic accidents in USA (Traffic Safety Facts 2013). In developing countries, due to issues
such as aggressive behaviours, lack of appropriate safety facilities and poor enforcement, the rate of
crash is even higher (Pawar and Patil, 2015). These reports along with the consideration of
pedestrians’ vulnerability indicate the importance of studies conducted on pedestrians’ safety issues
and behaviours.

Intersections are places which have the high potential for vehicle–pedestrians conflicts. Crossings,
designed as channels for pedestrians to cross the streets, are normally near signalised intersections.
These crossings cannot force them to obey traffic rules, specifically in off peak hours. Due to various
reasons including personal and traffic-related issues, pedestrians, as a group of road users,
sometimes violate traffic rules, taking risks and put themselves and other road users in danger by
performing high-risk behaviours. The most effective behaviour in pedestrian crossing time is to select
the appropriate time to collision (TTC) relative to vehicles approaching in each lane. Not waiting for a
safe TTC by a pedestrian can be based on poor judgment or violations that finally may lead to a
crash. TTC is defined as the time necessary for a vehicle to arrive at the crossing pedestrian in the
moment a pedestrian is crossing (Interval between pedestrians' crossing and vehicles' crossing from
conflict area) (Koh, Wong and Chandrasekar, 2014).

Pedestrians also may be threatened by risks resulted from their own carelessness; thus, errors and
violation must be distinguished. Errors is an unwanted deviation from safe practices reflecting
inadequate skills (e.g. inexperience) or temporary adverse states (e.g. fatigue). On the other hand,
violation is an intentional deviation from safe practices (e.g. deliberately violating a red light) that
reflects a person’s motivation for ignoring safety rules (e.g. a trade-off between risk and time lost)
(Twisk et al., 2015). Several authors had studied pedestrian behaviours to analyse the significance of
personal characteristics on their crossings. Serag (2014) examined pedestrians’ crossing behaviours
at un-signalised intersections by using video recording data using binary logit model and linear
regression analysis. The results showed that younger pedestrians usually choose smaller traffic gaps.
Koh, Wong and Chandrasekar (2014) examined pedestrians’ behaviours at seven signalised
intersections by using video recording data. They found that pedestrians that move in group are more
cautious than those who are crossing the intersections by themselves. They also found that women
are more cautious than men. Alhajyaseen and Iryo-Asano (2016) analysed continuous pedestrian
speed profiles at five signalised crosswalks to investigate sudden behavioural changes of pedestrians
(sudden speed change) in Qatar. Empirical analysis showed that sudden acceleration events were
observed at the entrance points to the pedestrian-vehicle conflict area. The results of multinomial logit
model showed that the entering speed, necessary speed to finish crossing before the onset of the
pedestrian signal red phase, and crosswalk length have a significant impact on speed change
choices. Brosseau and colleagues (2012) examined the effects of traffic lights on risky passing of
pedestrians and concluded that factors including type of traffic light (e.g. traffic lights with countdown

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
timer), pedestrian group size and flow, gender, age and maximum waiting time are significant in
pedestrian behaviours.

Another group of studies had focused on crossing pedestrians’ gap acceptance. Koh and Wong
(2014) examined gap acceptance of violating pedestrians at signalised intersections using binary logit
model. They found that pedestrians do not always wait for all lanes to be empty and cross the
intersections based on rolling gap acceptance. They also found that pedestrians typically cross the
first half of the intersections more cautiously. In Turkey, Onelcin and Alver (2015) conducted a study
to determine gap acceptance for safe crossing. They video recorded six intersections for data
collection and analysed their data using a one-way analysis of variance. The interactions between
times of crossing and positions of vehicles indicated that pedestrians crossed the intersections more
safely and with no desire to run when vehicles were at a distance of more than 50 metres. They also
reported that the average crossing time was almost two times longer when vehicles were at a
distance of 50-70 metres compared to when vehicles were at a distance of 25 metres.

Table 1 shows some of the studies in the field of risky behaviours of pedestrians crossing. As it can
be concluded, previous studies have been mainly conducted in signalised or un-signalised
intersections. Some common characteristics of pedestrians e.g. age, gender, distance and speed limit
have been chosen as factors to research whether they have an effect on pedestrians’ risk taking
behaviour or not. The main contribution of this study is defining the term of risk by crossing
pedestrians. Using stopping sight distance and reaction time of drivers as the basis for estimating the
risk term is a concept which was used in current study. In the case of pedestrian factors, clothing type
has been considered which has drawn less attention in recent studies. A common phenomenon in
Iranian cities is the clothing type of female pedestrians (Islamic dressing) which make them particular
compared to other countries. By identifying these factors, it will be possible to define indices for the
identification of risky and insecure areas for pedestrians that can be used in studies related to
pedestrians’ safety planning. Required data was gathered using video images in both signalised and
un-signalised intersections. Different characteristics of pedestrians' crossing behaviour has been
analysed and significant parameters have been reported in this study.

2. Data
To investigate factors affecting pedestrians’ risk taking behaviours, six different intersections were
selected in the city of Qazvin. Geographically, the selected areas were crowded intersections without
countdown signals. Features of each intersection are presented in Table 2 and Figures 1 and 2 show
a view of signalised and un-signalised intersections in this study. For data collection, pedestrians’
crossing behaviours during morning and evening peak hours (8-9:30 AM and 4-6 PM) were recorded
using video cameras mounted for surveillance at intersections. Video coverage included pedestrians’
waiting areas at each end of pedestrian crossing, length of crosswalk and the traffic light. The
selected days for filming were three consecutive mid-week days (Sunday, Monday and Tuesday) in
October 2015. Each video recording was continued for at least 90 minutes without interruption.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
In this study, risk takers are pedestrians who accept the time to collision (TTC) less than 3 seconds
(resulting from the stopping sight distance (SSD) from equation 1).

SSD = 0.278vt + 0.039

1.

Where,
: Speed of vehicle
t: Perception-reaction time
a: Decreasing acceleration

The SSD is the distance required for a vehicle to stop after seeing a pedestrian. The SSD is
composed of two parts: the perception-reaction time (PRT) (the time it takes for a driver to realise that
a reaction is needed; e.g. taking the foot off the accelerator and pressing the brake) and the
manoeuvre time (the time it takes to complete the reaction; e.g. to fully stop the car). Due to the low
speed of vehicles in the examined intersections in the present study (speeds below 50 km/h), the
manoeuvre time was considered 0.5 second and the PRT was 2.5 seconds; thus, 3 seconds was
considered as the time required for the drivers to react or for the pedestrians to select the TTC to
approaching vehicles. In this regard, some studies have estimated the critical gap or TTC by other
methods. For example, Pawar and Patil (2016) focuses on estimating pedestrians’ spatial and
temporal critical gaps at uncontrolled mid-block street crossings using both deterministic (Raff’s and
Ashworth’s method) and probabilistic approaches (Maximum Likelihood method and Logit method).
The results showed that temporal and spatial critical gaps by different methods vary between 3.6–4.3
s and 60–73 m respectively.

For the modelling of pedestrian behaviour, it is important to analyse the actions of pedestrians when
multiple vehicles are approaching the crossing in different lanes. In such a case, pedestrian has to
consider more than one vehicle for making gap acceptance decision. There are two crossing type
possibilities: (a) one-stage crossing (pedestrian waits until all lanes are empty and then crosses the
road) and (b) sequential crossings (the pedestrian will consider the gaps with reference to vehicles on
different lanes and accept rolling gaps). A pedestrian waiting at refuge area intended to cross [see to
Figures 3]. The positions of conflicting vehicles, A, B, C and D are as shown in the Figure 3. Vehicle A
is laterally nearest to the pedestrian and offers sufficient gap, but vehicle B is longitudinally very close
and the available gap is not sufficient to manoeuvre. Vehicle C is following B, and is sufficiently far
such that the gap between C and B is sufficiently large for a pedestrian to cross the approach safely.
Under such circumstances, it was observed that a pedestrian starts crossing the approach
anticipating safe gap, between vehicle B and C; By the time it crosses the path of vehicle A, vehicle B
has already cleared the crossing. Since vehicle C and D were sufficiently away, the pedestrian
crosses the path of vehicle C and D and reaches the other side of the road safely. This behaviour is
called rolling gap acceptance (Pawar and Patil, 2015).

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
In this study, variables were categorised in three groups of individual characteristics, environmental
and traffic conditions (Table 3). Some descriptive information of data is presented in Table 4.
Figures 4 and 5 show the frequency of pedestrians crossing signalised and un-signalised
intersections and number of risk taking pedestrians in terms of gender. Statistical analyses show that
women have less tendency for risk taking compared to men in signalised (45% versus 57%) and un-
signalised (33% versus 52%) intersections. It can also be concluded that men choose less TTC while
crossing at intersections. The results indicated that the average selected TTC at signalised
intersections were about 6.2 seconds for men and 7.1 seconds for women and at un-signalised
intersections they were about 5.6 seconds for men and 6 seconds for women. Analysis of pedestrians’
speed showed that men passed signalised and un-signalised intersections at average speeds of 1.5
m/s and 1.32 m/s and women passed them at average speeds of 1.3 m/s and 1.2 m/s. In other words,
both men and women passed signalised intersections at higher speed. In whole samples in the
present study, 55% of pedestrians carrying a bag, 72% of pedestrians using mobile phone and 14% of
pedestrians accompanied by a child did not select a safe TTC while crossing the intersections. These
findings indicate that the probability of pedestrians’ risk taking is more for those who carry baggage or
something in their hands (due to fatigue). Pedestrians who used their mobile phones while crossing
the intersections were unable to select an appropriate TTC. Being accompanied by a child increased
the probability of safe crossing, probably due to the sense of responsibility.

3. Procedure
As stated above, the focus of the current study is to explore the risk taking behaviour of pedestrians
crossing the street in both signalised and un-signalised intersections. Therefore, ‘‘taking risk’’ variable
is considered as the dependent variable, while the remaining variables (the variables listed in the
Table 4) were examined as independent variables. Since the dependent variable is dichotomous (0:
taking risk, 1: not taking risk), the model is developed with the binary logit method. Users choose an
alternative with the most utility. For each alternative (x), the utility function is defined as a linear
relationship composed of variables and constant values. To determine the probability of choosing
each alternative (p(x)) (a number between zero and one), the Binary Logit (equation 2) was used. It is
noteworthy that the parameters α and β are variable coefficient and constant value respectively.

2.

When the initial model was estimated, the best evaluation of the model is to analyse the estimated
coefficients, their values and the determination of significance level for each parameter. But, to
compare the initial model with other possible models, goodness of fit ( , ) were used to evaluate
the model. Goodness of fit for the evaluated model (a value between zero and one) was defined
based on equation (3) (Ortuzar and Willumsen, 1994 ; Kanafani, 1983):

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
3.

When only the constant values were non-zero, the goodness of fit was calculated based on the
following equation:

4.

In equations (3) and (4), LL(β) is the log likelihood function for all parameters; LL(C) is the log
likelihood function only for constant values (market share); and LL(0) is the log likelihood for models
with zero parameters (equal share). All values are between zero and one and values closer to one
indicated better fitting of the model.

4. Results
4.1. Model estimation
Using the NLOGIT 4.0 software, the following utilities were obtained for modeling pedestrians’ risk
taking when crossing intersections in Qazvin city. Results of the Binary Logit Model for signalised
intersections and goodness of fit of model are presented in Table 5.

Utility function of risk taking pedestrians in the binary logit model for signalised intersections is
indicated in (equation 5):
= 5.0751 – 2.0166 * PWB – 1.2005 * PLTTF – 1.0965 * FPWM + 0.9113 * PS – 0.2709 * TTC1 –
0.4392 * TTC2 – 0.0776 * SAV1 + 0.8205 * VP2 + 0.7381 * CP
5.

 A significant constant value is in the utility function that indicates the absence of some affecting
factors in the utility function of options. It must be noted that models without constant values were
also made in the present study; however, models with constant value provided better results.
 The results indicate that pedestrians accompanied by a child were more cautious in selecting
appropriate TTCs; in other words, they were more cautious when crossing the intersections as
they wanted to protect their children.
 The other variable that had a negative impact on the model was pedestrian looking direction at the
beginning of crossing an intersection. The results showed that pedestrians who were looking at
traffic flows were more cautious than those who were looking downward or at people. Obviously, a
pedestrian can realise traffic condition better and guess a more appropriate TTC when s/he is
looking at traffic flow.
 Analysis of the combined effects of gender and dressing type indicated that women wearing
manteau had less risky behaviours when crossing the intersections. Results of many studies have
indicated that women are more cautious than men (Koh, Wong and Chandrasekar, 2014; Tiwari et
al., 2007). Women wearing manteau have more freedom of action and can better analyse traffic
condition to select a more appropriate TTC. The interesting point is that in model without a

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
constant value, women wearing chador were introduced as higher risk takers, which was
consistent with this result.
 The variable of pedestrian speed had a positive impact on the model; accordingly, pedestrians
who were walking at higher speeds were less cautious. Probably, fast-moving pedestrians think
that they can easily pass shorter TTCs.
 Considering the definition of response variable and the concept of risk taking, with longer TTCs in
the first and second lanes, pedestrians’ risky behaviours decreased.
 Speed of approaching vehicle in the first lane had a significant negative effect in the model. It must
be noted that the first lane is defined based on the direction a pedestrian enters the crosswalk.
When the speed of approaching vehicle in the first lane was higher, the pedestrians’ tendency to
take risk decreased and they would select a longer TTC acceptance.
 In cases of more than one pedestrian violating the red light, other pedestrians’ tendency for risk
taking also increased. Pedestrians may think that drivers yield more easily if there are other
violating pedestrians.
 Kerb parking was another variable that positively affected pedestrians’ tendency to take risk. Kerb
parking decreases pedestrians’ sight, so that they cannot see approaching cars easily and may
select risky TTCs. On the other hand, kerb parking makes the length of pedestrian crossings
shorter, so that pedestrians may select shorter temporal gaps. These factors show that prohibiting
kerb parking at intersections is necessary.

Result of the binary logit model for un-signalised intersections and goodness of fit of model are
presented in Table 6. Equation 6 also shows the utility function of risk taking pedestrians for un-
signalised intersections:

= 3.0115 – 1.4476 * PWB + 4.4982 * RWC + 0.929 * PG – 1.4787 * OP – 0.0954 * WT – 0.3125 *

TTC1 – 0.3979 * TTC2
6.

From this equation, it can be concluded that:

 Pedestrians accompanied by a child were more cautious when crossing un-signalised
intersections; in other words, they selected a more appropriate TTC to pass the intersection more
securely.
 Running pedestrians selected shorter TTCs and were higher risk takers. These pedestrians may
think that they can easily pass the intersections by increasing their speed. The variable of running
pedestrian in this model used for un-signalised intersections was the equivalent of variable of
pedestrian speed in the model designed for signalised intersections.
 Old pedestrians are normally more cautious in selecting gaps or TTCs. These pedestrians may be
aware of their inability to react fast; they become more cautious when fast physical activities are
needed.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
 Waiting time had a negative impact on the model. In other words, when pedestrians waited for
longer periods of time before crossing the intersections, they also selected safer TTCs and passed
more safely.
 Similar to signalised intersections, the probability of pedestrians’ risk taking behaviours decreased
by increasing gaps or TTCs in the first and second lanes.
4.2. Discussion
In order to show clearly to what extent each of the independent variables affects the dependent
variable, an analysis of elasticities is carried out, as shown in Tables 7 and 8. The elasticity analysis
shows that, among the different variables, TTC1 (Time to collision in lane 1) and TTC2 (Time to
collision in lane 2) are the most influencing variables on pedestrians' risk taking behaviour crossing as
they have the highest elasticities. PWB (pedestrian with a baby) variable at signalised and un-
signalised intersections has the lowest effect.
Also, Figures 6 and 7 show the sensitivity analysis of pedestrians risk taking probability versus to
continuous variables such as pedestrian speed, time to collision, speed of approaching vehicle and
waiting time. Figure 6(a) shows that by increasing pedestrian crossing speed in signalised
intersections, pedestrians’ risk taking probability also increases. Figure 6(b), 6(c),7(b) and 7(c) show
the effects of time to collision in lane 1 and 2 on pedestrians’ risk taking probability in both signalised
and un-signalised intersections. These graphs have similar interpretation and express that by
increasing time to collision, trend of risk probability for pedestrians is decreasing in both types of
intersections. Sensitivity analysis of speed of approaching vehicle is pedestrians in Figure 6(d). This
graph shows that the probability of risk taking by pedestrians reduces if the approaching vehicle
speed increases. Also, by increasing waiting time variable in un-signalised intersections, pedestrians’
risk taking probability reduces.

5. Conclusion
The current paper aims to investigate the significance of various parameters affecting the risk taking
behaviour of pedestrians. Using video recording from 6 intersections in Qazvin, required data,
including 800 records were extracted. The pedestrians’ risk taking probability was modelled using
binary logit model and was analysed using elasticity analysis and sensitivity analysis. The overall
results indicated that factors affecting pedestrians’ risk taking behaviours can be categorised in three
groups of individual characteristics, environmental conditions and traffic conditions.

The most important factors affecting pedestrians’ risk taking in group of individual characteristics:

 Gender was a significant factor in pedestrians’ risk taking; men take more risks than women;
 Older pedestrians are more cautious than younger ones when crossing the intersections; they
mostly selected longer gaps;
 Women wearing manteau crossing the intersections more cautiously than women wearing chador;
 Pedestrians who walked faster selected smaller TTCs when crossing the intersections;

The most important factors affecting pedestrians’ risk taking in group of environmental conditions:

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
 Pedestrians accompanied by children were less likely to put themselves in danger;
 The violations of other pedestrians in the opposite direction positively affect other pedestrians’
tendency to risk;
 kerb parking negatively affected pedestrians’ sights and was considered as a risk factor;
 Longer waiting times for selecting appropriate gaps or TTCs led to less risk taking behaviours by
pedestrians;

The most important factors affecting pedestrians’ risk taking in group of traffic conditions:

 When a vehicle’s speed in the first lane is higher, pedestrians accept smaller gaps or TTCs;

By increasing time to collision, speed of approaching vehicle and waiting time, risk probability for
pedestrians is decreasing, but by increasing pedestrian speed, pedestrians’ risk taking probability
increases.

According to the results, providing training programs to increase safety awareness of different groups
of pedestrians seems quite necessary in order to prevent the occurrence of pedestrian related
conflicts and accidents. Although for having a safe pedestrian environment, many factors should be
considered but educational and public awareness campaigns can be helpful to promote safe crossing
behaviours among pedestrians.

6. Acknowledgement

The authors would like to address special thanks to Qazvin traffic control center for their contribution
and providing parts of the necessary data for this study.

7. Reference
Alhajyaseen, W.K. and Iryo-Asano, M., (2016). Studying critical pedestrian behavioural

changes for the safety assessment at signalised crosswalks. Safety science, 91,

pp.351-360.

Brosseau, M., Saunier, N., Le Mouel, K. and Miranda-Moreno, L., (2012). The impact of

traffic lights on dangerous pedestrian crossings and violations: a case study in

Montreal. In Transportation Research Board 91st Annual Meeting (No. 12-0941).

de Dios Ortúzar, J. and Willumsen, L.G., (1994). Modelling transport. New Jersey: Wiley.

Kanafani, A., (1983). Transportation demand analysis. Mc.Graw-Hill.Inc.

Koh, P.P. and Wong, Y.D., (2014). Gap acceptance of violators at signalised pedestrian

crossings. Accident Analysis & Prevention, 62, pp.178-185.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Koh, P.P., Wong, Y.D. and Chandrasekar, P., (2014). Safety evaluation of pedestrian

behaviour and violations at signalised pedestrian crossings. Safety science, 70,

pp.143-152.

Kröyer, H.R., (2015). Is 30km/ha ‘safe’ speed? Injury severity of pedestrians struck by a

vehicle and the relation to travel speed and age. IATSS research, 39(1), pp.42-50.

NHTSA, Traffic Safety Facts., (2013) Data: Pedestrians, Report Number: DOT-HS-812-124,

National highway traffic safety administration, Washington DC, USA, 2015.

Onelcin, P. and Alver, Y., (2015). Illegal crossing behaviour of pedestrians at signalised

intersections: factors affecting the gap acceptance. Transportation research part F:

traffic psychology and behaviour, 31, pp.124-132.

Pawar, D.S. and Patil, G.R., (2015). Pedestrian temporal and spatial gap acceptance at mid-

block street crossing in developing world. Journal of safety research, 52, pp.39-46.

Pawar, D.S. and Patil, G.R., (2016). Critical gap estimation for pedestrians at uncontrolled

mid-block crossings on high-speed arterials. Safety science, 86, pp.295-303.

Serag, M.S., (2014). Modelling pedestrian road crossing at uncontrolled mid-block locations

in developing countries. International Journal of Civil and Structural Engineering,

4(3), p.274.

Sun, R., Zhuang, X., Wu, C., Zhao, G. and Zhang, K., (2015). The estimation of vehicle

speed and stopping distance by pedestrians crossing streets in a naturalistic traffic

environment. Transportation research part F: traffic psychology and behaviour, 30,

pp.97-106.

Tiwari, G., Bangdiwala, S., Saraswat, A. and Gaurav, S., (2007). Survival analysis:

Pedestrian risk exposure at signalised intersections. Transportation research part F:

traffic psychology and behaviour, 10(2), pp.77-89.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Twisk, D.A., Commandeur, J.J., Vlakveld, W.P., Shope, J.T. and Kok, G., (2015).

Relationships amongst psychological determinants, risk behaviour, and road crashes

of young adolescent pedestrians and cyclists: Implications for road safety education

programmes. Transportation Research Part F: Traffic Psychology and Behaviour, 30,

pp.45-56.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Table 1. Some of the studies in the field of risky behaviours of pedestrians crossing
Type of
Number Authors Year Purposes Results
analyses
Pedestrians pass more safely in the
Examine risky situations for
Tiwari and Survival first half of an intersection. Average
1 2007 pedestrians at signalised
colleagues analysis waiting time for female pedestrians
intersections
is 27% more than male ones.
Pedestrian group size and flow,
Examine the effects of traffic
Brosseau and Logistic gender, age and maximum waiting
2 2012 lights on risky crossing of
colleagues regression time are significant in pedestrian
pedestrians
behaviours.
Younger pedestrians usually choose
Binary Examine pedestrians’ crossing
smaller traffic gaps; however, no
3 Serag 2014 Logit behaviours at unsignalised
significant difference exists between
intersections
young and middle-age groups.
Pedestrians do not always wait for all
lanes to be empty and cross the
Examine gap acceptance of
Koh and Binary intersections based on rolling gap
4 2014 violating pedestrians at
Wong Logit acceptance. Pedestrians typically
signalised intersections
cross the first half of the
intersections more cautiously.
Pedestrians that move in group are
Koh, Wong Examine pedestrians’ more cautious than those who are
Binary
5 and 2014 behaviours at signalised crossing the intersections by
Logit
Chandrasekar intersections themselves. They also found that
women are more cautious than men.
Injury severity of pedestrians
Over 30% of severe injury accidents
Multinomial struck by a vehicle and the
6 Kröyer 2014 occur in speed environments below
logit model relation to travel speed and
35 km/h.
age
Pedestrians crossed the intersections
Onelcin and Determine gap acceptance for more safely and with no desire to run
7 2015 ANOVA
Alver safe crossing when vehicles were at a distance of
more than 50 metres.
Determine pedestrian
Gaps acceptance were affected by
Pawar and Binary temporal and spatial gap
8 2015 size and speed of vehicle and
Patil Logit acceptance when crossing
pedestrians’ group size.
intersections
In sunny conditions, pedestrians
The estimation of vehicle
were able to accurately estimate
speed and stopping distance
Sun and Regression speeds that were lower than 40 km/h.
9 2015 by pedestrians crossing streets
colleagues equation In rainy conditions, pedestrians were
in a naturalistic traffic
able to accurately estimate speeds
environment
ranging from 35 km/h to 45 km/h.
The entering speed, necessary speed
Studying critical pedestrian
Alhajyaseen to finish crossing before the onset of
Multinomial behavioural changes for the
10 and 2016 the pedestrian signal red phase, and
logit model safety assessment at signalised
IryoAsano crosswalk length have a significant
crosswalks
impact on speed change choices.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Table 2. Features of intersections in this study
Average
Average
Number Length of Average Average Vehicles
Pedestrian
Intersection of crosswalk Cycle green volume
volume (every
lanes (m) length (s) time (s) (every 15
15 minutes)
minutes)
Adl 3 18.6 90 50 154 230
Valiasr 3 21.3 130 30 123 488
Ferdosi 2 11.7 40 20 73 194
Shohada 2 7 100 30 50 234
Khayam 2 12.5 Un-signalised 165 368
Valiasr
2 10 Un-signalised 59 363
midblock

Table 3. Variables considered for modelling

Individual characteristics Environmental conditions Traffic conditions
Pedestrian with object (bag /
Crossing length (m) Waiting time (s)
mobile phone / children) in hand
Pedestrian with company (alone /
The number of lanes (two lane / Crossing type (one-stage
one company / more than one
three lane) crossing / rolling crossing)
company)
Pedestrian speed Day time (morning / evening) Time to collision (s)
Running pedestrian (at speeds of Speed of approaching
Kerb parking
more than 1.9 m/s) vehicle (km/h)
The direction of pedestrian look Pedestrian crossing / not Pedestrians volume (every
(ground / traffic / people) crossing zebra crosswalk 15 minutes)
Existence of violating
Vehicles volume (every 15
Gender (male, female) pedestrian in red time for
minutes)
crossing pedestrian
Pedestrian crossing direction The number of pedestrians in
Age (adolescent / young / old)
(near end / far end) waiting area
Pedestrian dressing type (suit /sport
1 2
/manteau / chador )
1
A dress outfit with long sleeves covering body from shoulder to knees or lower on legs.
2
a large piece of cloth that is wrapped around the head and upper body leaving only the face exposed, worn especially by
Muslim women.

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Table 4. Descriptive information of data
Type
Variable Frequency Minimum Maximum Average
variable
Baby Binary 41 0 1 0.05
Pedestrian with
Mobile Binary 18 0 1 0.02
object
Bag Binary 211 0.26
Pedestrian speed Continuous 800 0.7 3.5 1.3
Male Binary 433 0 1 0.54
Gender
Female Binary 367 0 1 0.46
Alone Binary 512 0 1 0.64
One Binary 1
205 0 0.26
Pedestrian with company
company More than
one Binary 83 0 1 0.1
company
Crosswalk Binary 509 0 1 0.64
Situation crossing
Other Binary 291 0 1 0.36
Adolescent Binary 61 0 1 0.08
Age Young Binary 686 0 1 0.85
Old Binary 53 0 1 0.07
Independent variables

Suit Binary 27 0 1 0.03

Pedestrian Sport Binary 406 0 1 0.51
dressing Manteau Binary 225 0 1 0.28
Chador Binary 142 0 1 0.18
Crossing length Continuous 6 7 21 13.5
Day time Morning Binary 394 0 1 0.49
Evening Binary 406 0 1 0.51
Existence of violating pedestrian
in red time for crossing Binary 257 0 1 0.64
pedestrian
Pedestrian Near end Binary 414 0 1 0.52
crossing direction Far end Binary 386 0 1 0.48
Waiting time (s) Continuous 800 0 50 1.6
One-stage
Binary 612 0 1 0.23
crossing
Crossing type
Rolling 1
Binary 188 0 0.77
crossing
Time to collision (s) Continuous 800 1.1 53.5 6.2
Speed of approaching vehicle Continuous
800 10 55 25.2
(km/h)
Pedestrians volume (every 15 Continuous
800 9 191 98
minutes)
The number of pedestrians in Continuous
800 94 530 132
waiting area
Dependent variable

Taking risk Binary 800 0 1 0.47

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Table 5. Results of the binary logit model at signalised intersections
Significance Standard
Variable Symbol Coefficient
level Error
- Pedestrian with a baby PWB -2.0166 0.0588 1.0672
- Pedestrians look toward traffic flow at
PLTTF -1.2005 0.0591 0.636
beginning of crossing
- Female pedestrian wearing manteau FPWM -1.0965 0.0034 0.3745
- Pedestrian speed PS 0.9113 0.0438 0.452
- TTC in lane 1 TTC1 -0.2709 0.0000 0.5848
- TTC in lane 2 TTC2 -0.4392 0.0000 0.5633
- Speed of approaching vehicle in lane 1 SAV1 -0.0776 0.0001 0.1994
- Existence of more than one violating
pedestrian crossing an intersection in VP2 0.8205 0.0316 0.3816
opposite direction
- Existence of kerbside parking CP 0.7381 0.0643 0.3993
- Constant 5.0751 0.0000 1.0378
log likelihood (equal share) LL(0) -277.25
log likelihood (market share) LL(C) -276.95
log likelihood (Model coefficients) LL(β) -137.97
2
ρ 0= 1- 0.502
2
ρ c= 1- 0.501

Table 6. Results of the binary logit model at un-signalised intersections

Significance
Variable Symbol Coefficient Standard Error
level
- Pedestrian with a baby PWB -1.4476 0.0218 0.6309
- Running while crossing RWC 4.4982 0.0000 1.053
- Pedestrian gender PG 0.0929 0.0011 0.2844
- Old pedestrian OP -1.4787 0.0098 0.5728
- Waiting time before crossing WT -0.0954 0.034 0.045
- TTC in lane 1 TTC1 -0.3125 0.0000 0.0493
- TTC in lane 2 TTC2 -0.3979 0.0000 0.06
- Constant 3.0115 0.0000 0.436
log likelihood (equal share) LL(0) -277.25
log likelihood (market share) LL(C) -273.32
log likelihood (Model coefficients) LL(β) -166.53
2
ρ 0= 1- 0.399

2
ρ c= 1- 0.39

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Table 7. Elasticity analysis of independent variables at signalised intersections
Independent variable Symbol Elasticity of logit model
Pedestrian with a baby PWB - 0.033
Pedestrians look toward traffic flow at
PLTTF - 0.715
beginning of crossing
Female pedestrian wearing manteau FPWM - 0.171
Pedestrian speed PS 0.82
TTC in lane 1 TTC1 - 1.058
TTC in lane 2 TTC2 - 1.717
Speed of approaching vehicle in lane 1 SAV1 - 0.981
Existence of more than one violating
pedestrian crossing an intersection in VP2 0.13
opposite direction
Existence of kerbside parking CP 0.135

Table 8. Elasticity analysis of independent variables at un-signalised intersections

Independent variable Symbol Elasticity of logit model
Pedestrian with a baby PWB - 0.044
Running while crossing RWC 0.145
Pedestrian gender PG 0.328
Old pedestrian OP - 0.051
Waiting time before crossing WT - 0.113
TTC in lane 1 TTC1 - 1.186
TTC in lane 2 TTC2 - 1.566

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Figure 1. View signalised intersections in this study

Figure 2. View of un-signalised intersections in this study

Figure 3. Classification of crossing type

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Figure 4. Frequency of pedestrians and risk taking pedestrians in terms of gender in signalised
intersections

Figure 5. Frequency of pedestrians and risk taking pedestrians in terms of gender in un-signalised
intersections

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.
 Accepted manuscript doi:
10.1680/jmuen.17.00005
Figure 6. Pedestrians’ risk taking probability based on the variables a) pedestrian speed, b) time to
collision in lane 1, c) time to collision in lane 2 and d) speed of approaching vehicle in lane 1 in
signalised intersections

Figure 7. Pedestrians’ risk taking probability based on the variables a) waiting time, b) time to collision
in lane 1 and c) time to collision in lane 2 in un-signalised intersections

Downloaded by [ Glasgow University Library] on [09/11/17]. Copyright © ICE Publishing, all rights reserved.