Accident severity analysis using ordered probit model

To reduce injuries in road crashes, better understanding is needed between the relationship of injury severity and risk factors. This study seeks to identify the contributing factors affecting crash severity with broad considerations of driver characteristics, roadway features, vehicle types, pedestrian characteristics and crash characteristics using an ordered probit model. It also explores how the interaction of these factors will affect accident severity risk. Three types of accidents were investigated: two-vehicle crashes, single vehicle crashes and pedestrian accidents. The reported crash data in Singapore from 1992 to 2001 were used to illustrate the process of parameter estimation. Several factors such as vehicle type, road type, collision type, location type, pedestrian age, time of day of accident occurrence were found to be significantly associated with injury severity. It was also found that injury severity decreases over time for the three types of accident investigated.     

Level crossing safety on East Japan Railway Company: Application of probabilistic risk assessment techniques

This paper describes the application of probabilistic risk assessment techniques to level crossing safety on JR East, the largest of the six private railroads in Japan. The risk of a level crossing accident was defined as the product of the accident rate and the expected consequences per accident. Rail traffic volume, road traffic volume, visibility of the crossing from the road, road gradient, width of the crossing and the type of safety devices at the crossing were shown to influence the accident rate and the collective risk. The mean accident rate at all crossings was 0.74 per million trains. The accident rate was 0.59 per million trains at crossings equipped with barriers, 1.25 at crossings equipped with warning bells and 0.76 at pedestrian crossings. Crossings equipped with obstacle detectors had a lower accident rate (0.12 per million trains) than crossings without detectors (0.43 per million trains). Crossings with visibility less than 20 m had a 50% higher mean accident rate than crossings with visibility greater than 20 m. As the number of tracks increased, the accident rate monotonically increased due to the increased accident exposure. Risk assessment techniques were applied to determine the efficacy of the various level crossing safety devices. In addition to upgrading the safety of crossings, the management techniques stressed the importance of education campaigns in warning the public about the dangers of illegal crossings.     

Geographical variation in traffic-related mortality and morbidity among pedestrians in Quebec, 1983-1988

In this study we analyze the dmographic and geographical distribution of pedestrian accidents (n=28,452) in the Quebec Municipalités régionales de comté (MRC) reported between January 1, 1983 and December 31, 1988. In order to be able to make comparisons between MRCs, we have created in Comparative Mortality Index and a Comparative Morbidity Index. We use a logistic regression model to single out explanatory factors. The main tendencies which are apparent in the demographic analysis are an overrepresentation of young persons and elderly persons in the accident statistics, and that men are more at risk than women. Men 15 years and under and men 65 years and over are more at risk of being involved in a fatal pedestrian accident. The geographic analysis show that among men, the average mortality rate is one and a half times greater in rural MRCs than in urban MRCs; among women, rural mortality is two and a half times greater than urban mortality. In the case of non-severe injuries we note an over-representation of urban MRCs. Young persons 15 years or less who live in an urban, populated zone are those most at risk of being involved in a pedestrian accident. Identifying high risk groups or areas is a prerequisite step for injury prevention and control.     

Bus accidents,bus deregulation and London

Local bus services were deregulated in October 1986 in all areas of Britain except London. Government policy is to extend deregulation to London, though not in the current parliament. This paper analyses statistics on bus accidents from the national road accident database from 1981 to 1991 to compare results for London and the rest of Great Britain, and to consider whether deregulation has affected safety. The conclusions depend on the assumption that accident recording practice was not itself affected by deregulation.Bus accident rates are higher in London than on built-up roads elsewhere, partly apparently because of road traffic conditions in London, and partly because open-platform buses have higher accident rates involving occupants, including boarding and alighting accidents, than buses with doors.The main safety effects of deregulation operate through its effects on bus activity, though there is also some evidence that the rate per bus-kilometre of accidents involving other road users fell slightly. This may be due in part to the trend towards smaller buses associated with deregulation. Deregulation has led to a fall in bus patronage, and thus to a fall in occupant casualties; and to a rise in bus-kilometres, and thus to a rise in casualties among other road users in accidents involving buses. The number of fatal and serious casualties among other road users involved in bus accidents is larger than the number among bus occupants, so deregulation in London could lead on balance to a small rise in fatalities involving buses. On the other hand, the number of slight casualties among other road users involved in bus accidents is smaller than the number among bus occupants, so deregulation in London could lead to a fall in the number of slight casualties involving buses.Although there were fears that changes in the management of bus operations or financial pressures might lead to increased accident rates following deregulation, there no evidence in the findings to support such fears.This paper was first presented at the 3rd International Conference on Competition and Ownership in Surface Public Transport at Mississauga, Canada on 25–29 September 1993. The author is grateful to the Department of Transport for the speed and efficiency with which they provided special tabulations from the road accident database for this study.     

Accident cost saving and highway attributes

Two semi-logarithmic regression models are developed to estimate accident rates and accident costs, respectively, for rural non-interstate highways in the state of Iowa. Data on 21,224 accidents occurring between 1989 and 1991 on 17,767 road segments are used in the analysis. Seven road attributes of these road segments are included as predictor variables. Applying the resulting regression models to a rather typical highway upgrade situation, the present value of the accident cost saving is computed. The sensitivity of the estimated cost saving to values for fatal, personal injury, and property damage only accidents is tested.Because factors other than road characteristics greatly influence accident costs, the models developed in this research explain a limited amount of the variance in these costs among road segments. Results of the analysis indicate that the most important attribute associated with accident costs is average daily traffic per lane, followed by conditions requiring passing restrictions and the sharpness of curves. Varying the values for the three categories of accidents shows that results are far more sensitive to the value of personal injuries than fatalities. The feasibility of using predictive models of accident costs in benefit-cost analyses of highway investments is demonstrated.     

Assessing the long term benefit of banning the use of hand-held wireless devices while driving

An increasing number of legislative efforts have been undertaken to prohibit the use of hand-held wireless devices while driving. As of July 2012, ten states and the District of Columbia enforce laws banning the use of hand-held cell phones while driving. Thirty-nine states and the District of Columbia have banned text messaging while driving. Recent studies of driver behavior suggest that hand-held wireless device usage negatively impacts driver performance. However few studies at the aggregate level address the plausible link between the use of hand-held wireless devices while driving, increased risk of automobile accidents, and government legislative efforts to reduce such risk. This paper analyzes data at the aggregate level and builds a regression model to estimate the long term accident rate reduction due to a hand-held ban. This model differs from previous studies, which consider short term accident rate reduction, by considering time trends in the accident rate due to the ban. Additionally, counties considered in this analysis are placed into groups based on driver density, defined by the number of licensed drivers per centerline mile of roadway, and a separate analysis is performed within these groups. This approach allows one to better quantify the effect of hand-held bans in counties of different driver densities. Results from this paper suggest that bans on hand-held wireless device use while driving reduce the rate of personal injury accidents in counties with high levels of driver density, but may increase accident rates in counties with low driver density levels. These results can inform transportation policymakers interested in reducing automobile-accident-risk attributable to the use of hand-held wireless devices while driving.     

Valuing the Prevention of Road Accidents in Belgium

Abstract

In a recent international comparison of the social costs of road accidents, Trawén et al. (2003) noted that cost data are not available for Belgium and, by consequence, play no part in Belgian policy‐making. The purpose of the present paper is, therefore, to value the costs per casualty type and per accident in Belgium. Empirical data are provided on human and economic production losses as well as on direct accident costs such as medical costs, hospital visiting costs, accelerated funeral costs, property damage, administrative costs of insurance companies, litigation costs, police and fire department costs, and congestion costs. In Belgium the marginal unit value of preventing a road casualty is estimated at €2 004 799 per fatal casualty, €725 512 per seriously injured and €20 943 per slightly injured victim. The unit cost per accident amounts to €2 355 763, €850 033, €34 944 and €2571 for fatal, serious, slight injury and property damage only accidents, respectively. These results are consistent with valuations reported in other high‐income countries. Finally, the total costs of road accidents in 2002 are valued at €7.2 billion (2004 prices), or 2.6% of gross domestic product.     

HMB Reflectors: A New Horizontal Retro‐reflecting System to Improve Safety at Pedestrian Crossings

Abstract

Each year more than 1000 pedestrians are injured in accidents on pedestrian crossings in Switzerland. The accidents often occur in darkness, twilight or poor visibility during rain at locations without sufficient public street lighting because vehicle drivers notice the pedestrian crossing too late or overlook it altogether. Pedestrian crossings can be made significantly easier for vehicle drivers to recognize at night and in poor visibility by means of HMB reflectors. When crossing sites are made more conspicuous with high horizontal retro‐reflecting markers, the readiness to stop increases. The reflectors can thus contribute to improving road safety at pedestrian crossings. This new low‐cost measure has a wide range of applications. The new reflector system is currently gaining ground in Switzerland and several other European countries.     

On the identification of the effect of prohibiting hand-held cell phone use while driving: Comment

In a paper recently published in this journal (Nikolaev, A.G., Robbins, M.J., Jacobson, S.H., 2010. Evaluating the impact of legislation prohibiting hand-held cell phone use while driving. Transportation Research Part A 44, 182–193.), Nikolaev et al. (2010) provide evidences on the effect of hand-held cell phone bans on driving safety. More specifically, they analyze the impact of a state-wide ban on hand-held cell phone use while driving on the number of fatal automobile and personal injury accidents per 100,000 licensed drivers per year and conclude that the ban had a significant negative impact on both the mean fatal accident rate and the mean personal injury accident rate. In this paper I argue that they lack of a good identification strategy that enables them to correctly identify the causal effect of the ban. I also provide evidence that the effect they find is a combination of the ban effect and of unobservable variables not accounted for in their analysis. Finally, I provide a way where one can control for unobservables when estimating the causal effect of the ban and find that indeed that ban appears to have a negative effect on fatal automobile accidents.     

The effect of complex models of externalities on estimated optimal tolls

Transport externalities such as costs of emissions and accidents are increasingly being used within appraisal and optimisation frameworks alongside the more traditional congestion analysis to set optimal transport policies. Models of externalities and costs of externalities may be implemented by a simple constant cost per vehicle-km approach or by more complex flow and speed dependent approaches. This paper investigates the impact of using both simple and more complex models of CO₂ emissions and cost of accidents on the optimal toll for car use and upon resulting welfare levels. The approach adopted is to use a single link model with a technical approach to the representation of the speed-flow relationship as this reflects common modelling practice. It is shown that using a more complex model of CO₂ emitted increases the optimal toll significantly compared to using a fixed cost approach while reducing CO₂ emitted only marginally. A number of accident models are used and the impact on tolls is shown to depend upon the assumptions made. Where speed effects are included in the accident model, accident costs can increase compared to the no toll equilibrium and so tolls should in this case be reduced compared to the congestion optimal toll. Finally it is shown that the effect of adding variable CO₂ emission models along with a fixed cost per vehicle-km for accidents can increase the optimal toll by 44% while increasing the true welfare gained by only 8%. The results clearly demonstrate that model assumptions for externalities can have a significant impact on the resulting policies and in the case of accidents the policies can be reversed.

Updating predictive accident models of modern rural single carriageway A-roads

Abstract

Reliable predictive accident models (PAMs) are essential to design and maintain safe road networks, and yet the models most commonly used in the UK were derived using data collected 20 to 30 years ago. Given that the national personal injury accident total fell by some 30% in the last 25 years, while road traffic increased by over 60%, significant errors in scheme appraisal and evaluation based on the models currently in use seem inevitable. In this paper, the temporal transferability of PAMs for modern rural single carriageway A-roads is investigated, and their predictive performance is evaluated against a recent data set. Despite the age of these models, the PAMs for predicting the total accidents provide a remarkably good fit to recent data and these are more accurate than models where accidents are disaggregated by type. The performance of the models can be improved by calibrating them against recent data.     

Competing risks analysis on traffic accident duration time

Different clearance methods in traffic accident management lead to varied duration distributions. Apart from investigating the influence of various factors associated with accidents on the duration of such accidents using different clearance methods, this study also examines the cumulative incidence probability. We used traffic accident data obtained for 12 months from the Fourth Ring Expressway main line in Beijing to develop a subdistribution hazard regression model, which can assess the risk factors of two clearance methods. The regression results show that the different factors have statistically significant effects on the duration of two accident groups with different clearance methods; furthermore, opposite effects occur even for some factors that have a strong effect on both accident groups. For example, an accident involving a taxi extends the duration time with clearance method 1; in comparison, the accident is shorter with clearance method 2. The predicted cumulative incidence curves of the two types of clearance methods are shown as examples, with stratification based on the influence factors (taxi involved, season). Finally, the Gray test of the cumulative incidence functions and the log‐rank test of the Kaplan–Meier estimates of the survival functions are compared, in order to demonstrate the importance of using proper methods for analyses. Copyright © 2014 John Wiley & Sons, Ltd.     

全球民航事故调查数据统计与分析

为探索全球民航事故的发生及调查规律,从航空安全网(ASN)采集626起事故调查数据,对事故发生季度、类型、机型等特征进行统计分析,同时对事故调查的实施机构、调查时长、各类事故调查时长等特征展开分析。结果发现:全球民航事故随运行时间的增加呈现波浪形缓慢下降;事故数量最多的前两位为冲偏出跑道和飞机失去控制,并多集中在第三季度(7-9月);波音公司各机型事故曲线整体类似,同时与空客公司各机型飞机事故曲线存在差异。民航事故调查从开始到发布报告的平均周期为26.1个月;各调查机构针对冲偏出跑道、失去控制、可控飞行撞地事故的平均调查周期分别为22.2、40.4、12.4个月。研究结果为民航事故预防政策制定及事故调查提供参考依据。     

Estimating the value of risk reductions for car drivers when pedestrians are involved: a case study in Spain

We estimated the benefits associated with reducing fatal and severe injuries from traffic accidents using a stated choice experiment where choice situations were generated through a statistically efficient design. Specifically, the risk variables were defined as the expected annual number of vehicle car-users that suffered their death or were severely injured in a traffic accident. In addition, and differing from previous research, the number of pedestrians that died or were severely injured in traffic accidents per year was also included as a risk attribute in the choice experiment, to attempt at measuring drivers’ willingness to pay to reduce the risk of hitting pedestrians in a crash. The empirical setting was a choice of route for a particular trip that a sample of car drivers periodically undertakes in Tenerife, Spain. Models were estimated accounting for random taste heterogeneity and pseudo-panel data correlation. The median of the distribution of simulated parameters was used to obtain a representative measure for the monetary valuation of risk reductions. We found that the ratio between the values of reducing the risk of suffering a serious injury and that of reducing a fatality was approximately 18 %. Further, and quite novel, we also found that the value of reducing a pedestrian fatality was 39 % of the value of reducing a car occupant fatality.     

A novel variable selection method based on frequent pattern tree for real-time traffic accident risk prediction

With the availability of large volumes of real-time traffic flow data along with traffic accident information, there is a renewed interest in the development of models for the real-time prediction of traffic accident risk. One challenge, however, is that the available data are usually complex, noisy, and even misleading. This raises the question of how to select the most important explanatory variables to achieve an acceptable level of accuracy for real-time traffic accident risk prediction. To address this, the present paper proposes a novel Frequent Pattern tree (FP tree) based variable selection method. The method works by first identifying all the frequent patterns in the traffic accident dataset. Next, for each frequent pattern, we introduce a new metric, herein referred to as the Relative Object Purity Ratio (ROPR). The ROPR is then used to calculate the importance score of each explanatory variable which in turn can be used for ranking and selecting the variables that contribute most to explaining the accident patterns. To demonstrate the advantages of the proposed variable selection method, the study develops two traffic accident risk prediction models, based on accident data collected on interstate highway I-64 in Virginia, namely a k-nearest neighbor model and a Bayesian network. Prior to model development, two variable selection methods are utilized: (1) the FP tree based method proposed in this paper; and (2) the random forest method, a widely used variable selection method, which is used as the base case for comparison. The results show that the FP tree based accident risk prediction models perform better than the random forest based models, regardless of the type of prediction models (i.e. k-nearest neighbor or Bayesian network), the settings of their parameters, and the types of datasets used for model training and testing. The best model found is a FP tree based Bayesian network model that can predict 61.11% of accidents while having a false alarm rate of 38.16%. These results compare very favorably with other accident prediction models reported in the literature.     

Investigating effects of asphalt pavement conditions on traffic accidents in Tennessee based on the pavement management system (PMS)

Pavement maintenance is essential for ensuring good riding quality and avoiding traffic congestion, air pollution, and accidents. Improving road safety is one of the most important objectives for pavement management systems. This study utilized the Tennessee Pavement Management System (PMS) and Accident History Database (AHD) to investigate the relationship between accident frequency and pavement distress variables. Focusing on four urban interstates with asphalt pavements, divided median types, and 55 mph speed limits, 21 Negative Binomial Regression models were developed for predicting various types of traffic accident frequencies based on different pavement condition variables, including rut depth (RD), International Roughness Index (IRI), and Present Serviceability Index (PSI). The modeling results indicated that the RD models did not perform well, except for predicting accidents at night and accidents under rain weather conditions; whereas, IRI and PSI were always significant prediction variables in all types of accident models. Comparing the models goodness‐of‐fit results, it was found that the PSI models had a better performance in crash frequency prediction than the RD models and IRI models. This study suggests that the PSI accident prediction models should be considered as a comprehensive approach to integrate the highway safety factors into the pavement management system. Copyright © 2010 John Wiley & Sons, Ltd.     

The university workers’ willingness to pay for commuting

Using a dynamic approach, employing data on job mobility, we demonstrate that university workers?? marginal willingness to pay for reducing commuting distance is about ?0.25 per kilometre travelled. This corresponds to a marginal willingness to pay for reducing commuting time of about 75?% of the net average hourly wage. For females, the willingness to pay is substantially higher than for males. It is also substantially higher for workers that work few hours per day, as predicted by theory.     

Determinants of vessel-accident bunker spills

The study investigates determinants of vessel-accident bunker fuel spillage. A vessel-accident bunker spillage equation is estimated using Tobit regression and data of individual non-oil-cargo vessel accidents that were investigated by the US Coast Guard during 2001–2008. The results indicate that the bunker spillage of a freight ship and an offshore supply vessel accident will be greater than that of a passenger ship accident. Also, the bunker spillage of a non-oil-cargo vessel accident will be greater if the vessel accident is an abandonment and occurs at night, but less if the accident involves a vessel that has diesel propulsion and even less if the vessel has both diesel propulsion and a steel hull.     

Analysis of heavy truck accident severity

This paper presents a macro empirical study of injuries and severity, and their contributory factors, for heavy truck accidents. This analysis is carried out through the examination of the fatal injury and non-fatal injury odds ratio indices for a statistically best-fit, log-linear model. The statistical procedures used were the stepwise, backward elimination, log-linear modeling with a critical significance level of 0.05. The analysis was conducted on the set of accident severity contributory independent variables (main-effects) and all possible two-way interactions of these variables (interaction effects) for articulated heavy trucks. Among the findings of the study is that it would appear to be more effective to decrease the accident rates for these trucks rather than to try to limit their resulting severities in attempting to reduce the toll of heavy truck accidents.