A logistic model of the effects of roadway, environmental, vehicle, crash and driver characteristics on hit-and-run crashes

https://doi.org/10.1016/j.aap.2008.02.003Get rights and content

Abstract

Leaving the scene of a crash without reporting it is an offence in most countries and many studies have been devoted to improving ways to identify hit-and-run vehicles and the drivers involved. However, relatively few studies have been conducted on identifying factors that contribute to the decision to run after the crash. This study identifies the factors that are associated with the likelihood of hit-and-run crashes including driver characteristics, vehicle types, crash characteristics, roadway features and environmental characteristics. Using a logistic regression model to delineate hit-and-run crashes from nonhit-and-run crashes, this study found that drivers were more likely to run when crashes occurred at night, on a bridge and flyover, bend, straight road and near shop houses; involved two vehicles, two-wheel vehicles and vehicles from neighboring countries; and when the driver was a male, minority, and aged between 45 and 69. On the other hand, collisions involving right turn and U-turn maneuvers, and occurring on undivided roads were less likely to be hit-and-run crashes.

Introduction

Road crashes are a leading cause of death and injuries in many countries and extract a high cost on society. In the city-state of Singapore, for example, there were 9896 injuries in 2006 resulting from motor vehicle collisions, of which 190 were fatal (Singapore Police Force, 2007a). Among the different types of crashes, hit-and-run crashes are of interest to researchers because they are not only unethical acts but punishable offences as well. Also, leaving the victims at the crash scene delay crash notification and may result in an increase in the severity of the crash. Moreover, Lewis (1936) argued that victims of hit-and-run crashes were the most hapless group because of the reduced chance of getting compensation.

Hence, research aiming to improve the identification of hit-and-run vehicles and the drivers involved has been an area of interest in diversified fields such as medical, forensic, legal, insurance and engineering. Many studies in the field of medical science, for example, examined the types of injury sustained by victims in hit-and-run crashes to identify the types of vehicles involved. For example, Teresinski and Madro (2001) evaluated knee joint postmortem examinations of fatal pedestrian victims of traffic accidents to deduce the location as well as identify the type of vehicles involved. Karger et al. (2001) examined different types of fractures in the victims such as wedge-shaped bone fractures and fractures in the cervical and lumber spine to deduce the nature and the sequences of hit-and-run crashes.

The identification of vehicles involved in hit-and-run crashes is also an important task faced by forensic science laboratories. Debris which include headlamps, sidelight fragments, wheel track marks and paint fragments recovered at the accident location contribute to identifying the make and model of the offending vehicle. Several methods have been proposed to measure the color of retrieved paint fragments from crash scenes or from the clothing of victims to identify the vehicles involved in hit-and-run crashes (Cousins et al., 1989, Taylor et al., 1989, Locke et al., 1982, Locke et al., 1987, Locke et al., 1988).

Most of the studies discussed previously in various fields proposed different methods which would help to identify the fleeing vehicles in hit-and-run crashes. However, very few studies have explored the situations or circumstances under which hit-and-run crashes occurred. In one of the few studies that examined this issue, Solnick and Hemenway (1995) explored how victim's characteristics, driver's characteristics and circumstances of collision affected the hit-and-run choice in fatal pedestrian crashes. Their results showed that a driver was less likely to run when the victim was a child or an elderly pedestrian, if the crash occurred in the southern part of the America or in daylight, the driver was elderly, and the car driven was less than 5 years old. On the other hand, a driver was more likely to run when the victim was between 16 and 25 years old, the crash occurred in an urban area, on weekends, or in summer, and the driver was male or had no valid driver license, previous DWI convictions, positive or unknown BAC. In addition, Solnick and Hemenway (1994) found that 19% of all road crashes in the United States in 1989–1990 were hit-and-run cases caused by drunk driving and the hit-and-run motorists were disproportionately young and male. Although these studies gave some insight about the factors influencing drivers’ decision to leave the crash scenes, they were restricted to only pedestrian crashes.

In this study, a logistic model will be estimated to examine the influence of road features, vehicle attributes, environmental factors, crash characteristics and driver particulars on hit-and-run crashes in Singapore. This study will contribute to the literature in several ways. First, it will extend the study by Solnick and Hemenway (1995) by examining all hit-and-run crashes instead of pedestrian involved crashes alone. Second, it will examine several additional influences including the types of vehicles involved in the crashes, the vehicles’ countries of origin, types of crashes, vehicle maneuvers, ethnicities of drivers, presence of surveillance cameras and other location information like residential housing, retail or central business district areas, which are expected to be very relevant.

Last, it will examine hit-and-run crashes in an Asian country with a different social and cultural environment. Previous safety related studies in Singapore had focused mainly on examining intersection crashes (Chin and Quddus, 2003, Rifaat and Chin, 2007, Kumara and Chin, 2006, Mitra et al., 2002, Tay and Rifaat, 2007), motorcycle crashes (Quddus et al., 2002, Yuan, 2000), and red light running behaviors (Lum and Wong, 2002, Lum and Wong, 2003). No publicly available study has been found that deals with hit-and-run crashes in Singapore.

Singapore is a city-state located on a small island (700 km2) at the southern tip of the Malayan Peninsula in Southeast Asia. It has a total population of about 4.48 million, with a resident population of 3.61 million (Statistics Singapore, 2007).3 The actual composition of the population is quite fluid due to high percentage of foreign workers. The resident population comprises of four major ethnic groups: Chinese (75.2%), Malay (13.6%), Indian (8.8%) and Others (2.4%).4 The vast majority of the population stays in high-rise apartments due to the shortage of land. Singapore has one of the highest population densities in the world and is almost 100% urbanized. Therefore, although it has a very good quality road system and fairly high gross domestic product per capita (S$46,832),5 very stringent vehicle ownership and use policies are implemented to mitigate traffic congestion (McCarthy and Tay, 1993, Tay, 1996). Nevertheless, passenger cars (private, rental and taxis) still comprise the majority (62.3%) of the 799,373 registered vehicles, followed by goods vehicles (18.1%), motorcycles and scooters (17.9%) and buses (1.8%).

Section snippets

Analytical framework

The decision of a driver to stay and report the crash or to leave the scene without reporting the collision can be analyzed using the standard decision analysis framework based on the expected costs and benefits of the choices. There are few uncertainties associated with the outcomes of reporting a crash. The costs of the reporting crash consist mainly of an increase in the insurance premium for reporting the crash if the driver is at fault and potential penalties associated with any illegal

Methodology

In our study, the response variable, hit-and-run or non-hit-and-run crash, is a binary or dichotomous variable. Therefore, the logistic regression is a suitable technique to use because it is developed to predict a binary dependent variable as a function of predictor variables. The logistic regression model is widely used in road safety studies where the dependent variable is binary (Valent et al., 2002, Jones and Whitfield, 1988, Lui et al., 1988, Shibata and Fukuda, 1994, Zhang et al., 2000,

Discussion of results

The estimation results for the final model are shown in Table 2. Based on the p-values of the t-tests, 28 variables from 16 factors were found to be significant (p  0.05) or marginally significant (p  0.1). As suggested by Kockelman and Kweon (2002), variables with low statistical significance might also be retained in the model if they belonged to factors that had some significant effect on injury severity. Although this approach may reduce the efficiency of the estimates, it was adopted for

Conclusion

Although various measures such as improvements in vehicle safety and road designs as well as developments of telecommunication systems and increasing application of information technology in transportation systems have brought about better traffic safety standards, they have little impact in reducing hit-and-run crashes. Running after a crash is more likely to depend on the situational factors surrounding the crash. Although in-depth field investigations are preferred and should be considered

Acknowledgements

Support from the Natural Sciences and Engineering Research Council of Canada, the Alberta Motor Association and the Centre for Transportation Engineering and Planning are gratefully acknowledged. The authors also thank the Singapore Traffic Police for providing the data.

References (41)

Cited by (102)

View all citing articles on Scopus
1

Tel.: +1 403 220 5970; fax: +1 403 282 7026.

2

Tel.: +65 6874 2550; fax: +65 6779 1635.

View full text