Pedestrian red-light violations and other safety-related behaviours at signalised crosswalks

ABSTRACT

In this study, pedestrians’ violation rate at signalised crosswalks and the associated risk factors were investigated. The data for the study were obtained through roadside pedestrian observational surveys at 10 selected crosswalks within Accra Metropolis, Ghana. The associated risk factors for the red-light violations were determined using mixed-effect logistic regression model. The descriptive statistics revealed that the red-light violation rate in the Accra metropolis was approximately 62%. Most (63.5%) of the violations occurred during the evening and on weekends (73.6%). Over 98% of pedestrians demonstrated safety consciousness by way of crossing behaviour before and during crossing by observing oncoming traffic. From the mixed-effect logistics regression model, six independent variables being age, signal cycle length, number of pedestrians crossing at a time, day of the week and pedestrian light observation significantly influenced the risk of red-light violation by pedestrians. Effective law enforcement, education campaigns and engineering measures could be used to reduce the tendency of red-light violations by pedestrians and improve pedestrian safety at signalised crosswalks in the Accra metropolis.

KEYWORDS:

• Pedestrian
• red-light violation
• safety behaviors
• signalised crosswalk
• signal cycle length
• Ghana

1. Introduction

A Global Status Report on Road Safety by the World Health Organization (WHO) indicated that the deaths associated with road traffic crashes in 2021 was estimated to be 1.19 million (WHO, 2023). Pedestrians constitute 23% of the total deaths. The situation of pedestrian fatalities in Africa is relatively high and represents 27% of the total deaths from road traffic crashes in the subregion (WHO, 2023). Similar to many other low- and middle-income countries, Ghana has enormous challenges in terms of road safety. Pedestrian-vehicle collisions continue to be a public health concern in Ghana, accounting for a higher proportion (33.6%) of total road traffic fatalities in 2022 (NRSA, 2024). In the Accra Metropolis, pedestrians have the highest share (47.9%) of traffic-related deaths (NRSA, 2024).

In a mixed traffic environment, pedestrians are forced to share the road with other road users, and they bear higher responsibility in the event of a traffic crash (Bayomi et al., 2023; Demir et al., 2019). One of the road safety devices to reduce pedestrian-vehicle collisions on a road segment or intersection is traffic signals. Signalised pedestrian crosswalks provide vulnerable road users with a safe, convenient, and acceptable level of service, especially for individuals who have physical disabilities or reduced mobility and orientation. While signalised crossings with crosswalks seem to help with traffic management, they are unable to persuade pedestrians to follow the signals (Sisiopiku & Akin, 2003). Pedestrians’ reckless behaviour in failing to respect the red-light at signalised crosswalks has resulted in a huge number of traffic conflicts at intersections in both developed and developing countries, and this has resulted in a considerable number of deaths and serious injuries (Demir et al., 2019; Raoniar & Maurya, 2022).

Existing studies have shown that red-light violations are associated with various road and traffic characteristics, including vehicular traffic volume (Afshari et al., 2021; Shah & Pradhananga, 2024), signal cycle length (Jalali Khalilabadi et al., 2023; Mukherjee & Mitra, 2020; Raoniar & Maurya, 2022), waiting time (Haque & Kidwai, 2023), crossing length (Afshari et al., 2021; Shah & Pradhananga, 2024), crossing path and speed (Haque & Kidwai, 2023; Raoniar & Maurya, 2022), pedestrian group crossing (Anapali et al., 2021; Raoniar & Maurya, 2022; Shah & Pradhananga, 2024), pavement marking and on-street parking (Mukherjee & Mitra, 2020) and pedestrian demographic characteristics (Haque & Kidwai, 2023; Mukherjee & Mitra, 2020; Cambon de Lavalette et al., 2009; Guo et al., 2011; Hamed, 2001). Another most important factors contributing to illegal pedestrian crossings is the pedestrians’ sense of what it means to be safe (Chen et al., 2011).

Although several studies have been conducted on pedestrian crossing behaviour and red-light violations at signalized intersections, and the contributing factors. Most of these studies were conducted in the developed world where the situation is known to be low. Due to variations in road conditions and infrastructure, as well as pedestrian behaviours and socio-demographic characteristics among different countries, findings from one country may not be transferable to another. In Ghana, there is limited studies that uncover pedestrian red-light running rate and the associated risk factors. As part of the effort to improve the safety of signalised crosswalks in the urban environment in Ghana, it is necessary to understand the inappropriate pedestrian crossing behaviour as well as red-light violations at pedestrian crosswalks. This information will help to provide recommendations for planning, management, and control strategies, which will ultimately result in intelligent pedestrian crossing infrastructure systems on the roadways. Thus, this study contributes to the ongoing discussion on pedestrian safety by exploring the red-light violation rate and the associated risk factors among pedestrians in the Accra Metropolis, Ghana.

2. Materials and methods

2.1. Study area and data collection

The study was limited to the Accra Metropolis, Ghana in areas with signalized crosswalks. The choice of the study area was influenced by the high number of pedestrian crashes recorded for the region in general (NRSA, 2021). The data used in this study was obtained through an observational survey designed to determine red-light running rate among pedestrians and their safety-related behaviours in the Accra Metropolis, Ghana. The survey was undertaken at ten different locations within and outside the central business district (CBD) of the metropolis from 6^th July to 11^th August 2021 as shown in Figure 1. There were thirty-one (31) traffic lights with operable pedestrian signals at the study area, and the locations with the highest number of pedestrian volumes were selected for the study. At each location, the study took place for one weekday and a weekend. On a given day, the survey was conducted at two different times: morning (6:00–9:30 am) and evening (2:30–6:00 pm) to capture pedestrian characteristics for the morning and evening peak periods. In addition, the observations at each location were done at least twice a week (i.e. no less than one in a weekday and another on a weekend) on a non-rainy day. A day-long observation training session was held for observers prior to the data collection. At the training, the degree of consensus on various variables was measured. The inter-observer agreement reached 100% for all variables except the age group, where consensus still exceeded 95%”.

Figure 1. Map of the Accra Metropolis showing the survey locations.

During the survey, the location-specific characteristics recorded include signal cycle length (≤2 min/above 2 min), junction type (T-junction/crossroad/not at junction), presence of pedestrian refuge (present/absent), and posted speed limit (present/absent). The pedestrian specific characteristics and behaviour recorded include gender (male/female), age (less than 18 years/18 years and above) waiting location (curb/carriageway), light observation before crossing (yes/no), crossing pace (walking/running), crossing path (straight/diagonal) and violation status (violated/not violated).

2.2. Model specification

Following Nkrumah et al. (2022), the risk factors for pedestrian red-light violations were characterized using mixed-effect logistic regression model. The use of this model will help to account for possible heterogeneity due to variations in the survey locations. Let $y_{\mathit{ij}}\in \left\{0,1\right\}$ represent the outcome of the $i^{\mathit{th}}\left(\mathit{i}=1,2,\dots ,n_j\right)$ pedestrian crossing at signalized intersection at the $j^{\mathit{th}}\left(1,2,\dots ,\mathit{m}\right)$ location, the mixed-effect logistic regression model is defined as:

(1) $ln\left[\frac{{\mathit{\eta }}_{\mathit{ij}}}{1-{\eta }_{\mathit{ij}}}\right]=X_{ij}^T\mathit{\beta }+Z_j^T{\gamma }_j$(1)

where ${\eta }_{\mathit{ij}}=Pr\left(y_{\mathit{ij}}=1\right)$ represents the probability of the $i^{\mathit{th}}$ pedestrian violating red-light at the $j^{\mathit{th}}$ location, $X_{\mathit{ij}}\mathrm{and}{Z}_j$ represent vectors of fixed and random effect explanatory variables considered in the model with the associated unknown parameters being $\mathit{\beta }\mathrm{and}\mathit{\gamma },$ respectively. These unknown parameters in (1) were estimated by maximizing the log-likelihood function of (1) defined as:

(2) $log\mathit{L}\left(\mathit{\beta },\mathit{\gamma }\right)=\sum _{\mathit{i}=1}^n\sum _{\mathit{j}=1}^m\left[y_{\mathit{ij}}-1\right]\left[{\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\mathrm{T}}\mathit{\beta }{+\mathrm{Z}}_{\mathrm{j}}^{\mathrm{T}}{\gamma }_j\right]-log\left[1+exp\left(-{\mathrm{X}}_{\mathrm{i}\mathrm{j}}^{\mathrm{T}}\mathit{\beta }-{\mathrm{Z}}_{\mathrm{j}}^{\mathrm{T}}{\gamma }_j\right)\right]$(2)

The log-likelihood function is maximized by finding the first-order derivate and equating it to zero. However, the resultant equations have no closed form, so the adoptive Gaussian quadrature algorithms are used to obtain the estimates (Pinheiro & Chao, 2006). The random effect parameter $\mathit{\gamma }$ is assumed to have a normal distribution with mean 0 and variance ${\sigma }^2.$ The variance ${\sigma }^2$ component describe the degree of heterogeneity in the pedestrian red-light violation across different survey locations. If pedestrian red-light violations are homogeneous across the survey locations, then the mixed-effect logistic regression model will reduce to the standard logistic regression model. Based on Equation (1)(1) $ln\left[\frac{{\mathit{\eta }}_{\mathit{ij}}}{1-{\eta }_{\mathit{ij}}}\right]=X_{ij}^T\mathit{\beta }+Z_j^T{\gamma }_j$(1) , the predicted probability of the $i^{\mathit{th}}$ pedestrian violating red-light will be defined as:

(3) ${\hat{\eta }}_{\mathit{ij}}=\frac{exp\left(X_{ij}^T\hat{\beta }+Z_j^T{\hat{\gamma }}_{\mathit{j}}\right)}{1+exp\left(X_{ij}^T\hat{\beta }+Z_j^T{\hat{\gamma }}_{\mathit{j}}\right)}$(3)

The inclusion of a given explanatory variable in the model was based on 5% significance level. For categorical variables with more than two categories, the inclusion of such a variable in the model is based on the significance of at least one category. The performance of the competing models was compared using the Akaike information criterion (AIC) and the Bayesian information criterion (BIC).

3. Results

A total of 737 pedestrians were observed across all the selected sites within the Accra metropolis. Overall, the prevalence rate of red-light violations among the pedestrians was approximately 62% (Table 1). Majority (64.2%) of the violators observed were males. In addition, majority of the violators were 18 years and above. Pedestrian violations of red-light were highly (73.6%) observed on weekends compared to that of weekdays, and this usually occurred in the evening time of the day. Pedestrian refuge was present in most of the locations where a higher proportion of violation took place. Most (70.0%) of the pedestrians who violate the red-lights usually do not observe the light before crossing. In addition, they were also running while crossing the road. Most of the violations took place within the central business district of the study area, particularly in areas where the cycle length of the light is more than 2 min. The violation rate was observed to increase as the number of lanes increased. Thus, a 4-lane section of road recorded the highest violation rate of 65.1%, followed by a 3-lane section with a violation rate of 63.7% and the least of 59.8% for 2-lane sections (Figure 2). In addition, the violation was much less observed on non-junction type (road segments) of the road compared to crossroads or T-junctions.

Figure 2. Distribution of pedestrian red-light violations classified by (a) the number of lanes and (b) the type of junction.

Table 1. Descriptive statistics of the pedestrian red-light violations classified by other measured variables considered.

Variable	Frequency	Non-violations (%)	Violations (%)
Total observations	737 (100)	278 (37.72)	459 (62.28)
Gender
Male	350 (47.49)	126 (36.00)	224 (64.00)
Female	387 (52.51)	152 (39.28)	235 (60.72)
Age (years)
Less than 18	120 (16.28)	61 (50.83)	59 (49.17)
18 and above	617 (83.72)	217 (35.17)	400 (64.83)
Day of week
Weekday	438 (59.43)	199 (45.43)	239 (54.57)
Weekend	299 (40.57)	79 (26.72)	220 (73.58)
Time of day
Morning	413 (56.04)	161 (38.98)	252 (61.02)
Evening	324 (43.96)	117 (36.11)	207 (63.89)
Pedestrian refuge
Present	717 (97.29)	267 (37.24)	450 (62.76)
Not present	20 (2.71)	11 (55.00)	9 (45.00)
Posted speed limit
Present	445 (60.38)	151 (33.93)	294 (66.07)
Not present	292 (39.62)	127 (43.49)	165 (56.51)
Signal cycle length
2 min or less	298 (40.43)	131 (43.96)	167 (56.04)
More than 2 min	439 (59.57)	147 (33.49)	292 (66.51)
Waiting location
Curb	370 (99.05)	277 (37.95)	453 (62.05)
Carriageway	7 (0.95)	1 (14.29)	6 (85.71)
Light observation
Yes	264 (35.82)	129 (58.13)	135 (41.87)
No	473 (64.18)	149 (29.96)	324 (70.04)
Crossing pace
Walking	553 (75.03)	234 (42.31)	319 (57.69)
Running	184 (24.97)	44 (23.91)	140 (76.09)
Crossing path
Straight	723 (98.10)	276 (38.17)	447 (61.83)
Diagonal	14 (1.90)	2 (14.29)	12 (85.71)
Location type
Within CBD	437 (58.89)	153 (35.25)	281 (64.75)
Outside CDB	303 (41.11)	125 (41.25)	178 (58.75)

The contributing risk factors for the pedestrian red-light violations were established using mixed-effect logistic regression model. This model accounts for the unobservable variation due to the locations of the survey. The parameters and the associated p-values of the mixed-effect logistic regression model are summarized in Table 2. The inclusion of the variable in the model was decided based on 5% significance level. The variance of the location parameter was found to be statistically significant suggesting the use of the mixed-effect model as opposed to the standard logistic regression model. In addition, the mixed-effect model was found to outperform the standard fixed-effect logistic regression model based on the AIC, and BIC as shown in Table 3. Out of the 14 independent variables considered, only six of them were found to be statistically significant. These include the age of the pedestrian, day of the week, number of people crossing the road, circle length of the signal, crossing pace and light observation by the pedestrian.

Table 2. Estimated parameters for the mixed-effect logistic regression model for pedestrian red-light violation.

Variable	Coefficient	p-value	Odds Ratio	95% CI of OR
Age (Ref: Less than 18)
18 and above	0.467	0.035	1.595	1.035, 2.460
Day of week (Ref: Weekday)
Weekend	0.648	p<0.001	1.912	1.351, 2.705
Individuals crossing (Ref: Alone)
Not alone	−0.701	0.002	0.496	0.320, 0.769
Cycle length (Ref: 2 min or less)
More than 2 min	1.212	p<0.001	3.360	1.931, 5.853
Light observation (Ref: Yes)
No	0.669	p<0.001	1.952	1.395, 2.735
Crossing pace (Ref: Walking)
Runing	1.028	p<0.001	2.95	1.865, 4.191
Constant	−1.022	0.064
Location
variance	0.349	p<0.001
Goodness-of-fit Statistics
Number of observations	737
Log-likelihood at intercept	−438.666
Log-likelihood at convergence	−431.831
Wald χ^2 test statistic (p-value)	92 (p<0.001)

Table 3. Performance comparison between the fitted mixed-effect logistic regression model and the standard logistic regression model.

Statistic	Logistic regression	Mixed-effect logistic regression
AIC	893.07	879.66
BIC	925.29	916.48
Likelihood ratio test statistic	15.41
Likelihood ratio test p-value	p<0.001

4. Discussions

This study sought to determine the rate of red-light violation by pedestrians at signalised crosswalks in the Accra Metropolis. The observation results revealed that the red-light violation at pedestrian crosswalks in the study area was 62%. This rate seems to be on the higher side compared to rates recorded in other countries. For instance, Haque and Kidwai (2023) observed a lower red-light violation rate of 20% in New Delhi, India. Similarly, a violation rate of 5.5% was observed in Tel Aviv in a pedestrian crossing behaviour study (Rosenbloom 2009). The higher violation rate in this current study means that more pedestrians do not obey traffic rules in the capital city of Ghana, which is a recipe for pedestrian crashes. Again, this safety behaviour could be due to less punitive traffic offences in the system committed by road users. The relatively higher pedestrian red light violation rate revealed by the current study can also be explained as an indication of an unsatisfactory level of service in terms of signal timing and traffic levels within the Accra Metropolis at signalised crosswalks as argued by Koh et al. (2014). More pedestrians violated red-lights at locations where the cycle length was high.

The study results from the estimated model indicated that the propensity of red-light running rate by adult pedestrians was approximately 50% higher compared to young pedestrians (<18 years). However, this contradicts the findings of the existing literature as young people have been found to have bad road crossing behaviour as pedestrians (Ma et al., 2020). In addition, other studies have also documented that middle-aged and elderly are more cautious and law-abiding group who obey road safety regulations (Dommes et al., 2015; Ren et al., 2011). These contradictions may be due to differences in the cultural and environmental settings of the study areas. It could also be due to the fact that the National Road Safety Authority has been conducting road safety educational programmes in schools. The young people might be learning from these training programmes. In addition, the tendency for adults to violate red-light may be influenced by distractions such as mobile phone usage (Haque & Kidwai, 2023).

Again, pedestrians were about 90% more likely to violate red-light during weekends compared to weekdays, and this is inconsistent with existing research that weekends have no significant effect on red traffic violations for pedestrians (G. Zhang et al., 2016). The direction of the results may be influenced by the fact that pedestrians may find themselves in a more leisurely frame of mind during the weekend leading to less adherence to road safety rules. Additionally, there may be a higher number of intoxicated pedestrians during weekends due to social events and gathering leading to higher red-light violations (Hezaveh & Cherry, 2018).

The number of individuals crossing the road was found to influence red-light as has been observed in earlier studies (Anapali et al., 2021; Raoniar & Maurya, 2022; Shah & Pradhananga, 2024). The higher level of red-light compliance among pedestrians associated with group crossing is in line with earlier research works explaining that people crossing in groups seemed to obey traffic rules more than when pedestrians cross alone due to the influence of group power (Raoniar & Maurya, 2022). Additionally, crossing in a group may be safer, as there is a shared responsibility to adhere to traffic signals compared to individual crossing. The behaviour of a pedestrian is mostly influenced by the behaviour of others in the signalized intersection as argued by Anapali et al. (2021).

With regards to signal circle length, pedestrians were over 200% more likely to violate red-light in areas where signal cycle length was two minutes or more. Previous studies found that the signal cycle length significantly influences the signal violation behaviour by pedestrians (Raoniar & Maurya, 2022). Longer signal cycles may result in pedestrians attempting to cross against the signal, driven by a desire to minimize waiting times.

Pedestrians who did not look at the traffic light before crossing were 95% more likely to violate red-light compared to those who observed the signal before crossing the road. Similar to findings by Dommes et al. (2015), pedestrians are more likely to violate red-light when they cross without looking at the traffic light. This behavior might be influenced by the time pressure to reach destination faster (W. Zhang et al., 2016; Zhu et al., 2021) and distractions including mobile phone usage (Haque & Kidwai, 2023; Perra et al., 2022). According to Zhu et al. (2021), there is a trade-off between time and safety, and pedestrians are more sensitive to a reduction in time lost as compared to increase safety risk.

Similarly, pedestrians who run while crossing the road were about 195% more likely to violate the red-light while crossing the road compared to pedestrians who cross by walking this is in agreement with earlier research by Russo et al. (2018) who alluded that in contrast to pedestrians who adhered to traffic signals, those who crossed against the red light exhibited a higher propensity for running.

5. Conclusions and recommendations

This study has given insight into the prevalence and characteristics of red-light running by pedestrians at signalised crosswalks in the Accra Metropolis, Ghana. The result of this study has revealed the risk factors associated with pedestrian red-light violations include the age of the pedestrian, day of the week, number of people crossing the road, signal circle length, crossing pace and light observation by the pedestrian. The pedestrian red light violation rate for the city of Accra which was found to be 62% is on a higher side as compared to similar studies carried out in other cities. The results provided relevant information which could be useful in directing and supporting road safety policies and programs intended to improve road safety practices, particularly at signalised crosswalks.

Safety awareness of pedestrians is a key factor in improving safety at signalised crosswalks besides the improvement of infrastructure. With a higher level of red-light violations experienced in the Accra Metropolis, attention should be focused on obeying traffic rules and regulations by pedestrians and should be started from childhood by possibly stimulating pedestrian psychology and memory by videos of pictures that are resulting from illegal crossing crashes. Pedestrian waiting times at various junctions must be reduced in order to reduce the number of violations committed by pedestrians.

One limitation of the study is that it only considered traffic lights in the Accra Metropolis with operable pedestrian signal, potentially excluding valuable data from areas with traffic lights without such signals.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.