Important omitted spatial variables in safety models: Understanding contributing crash causes at intersections

Advances in safety research—trying to improve the collective understanding of motor vehicle crash causation—rests upon the pursuit of numerous lines of inquiry. The research community has focused on analytical methods development (negative binomial specifications, simultaneous equations, etc.), on better experimental designs (before-after studies, comparison sites, etc.), on improving exposure measures, and on model specification improvements (additive terms, non-linear relations, etc.). One might think of different lines of inquiry in terms of ‘low lying fruit’—areas of inquiry that might provide significant improvements in understanding crash causation. It is the contention of this research that omitted variable bias caused by the exclusion of important variables is an important line of inquiry in safety research. In particular, spatially related variables are often difficult to collect and omitted from crash models—but offer significant ability to better understand contributing factors to crashes. This study—believed to represent a unique contribution to the safety literature—develops and examines the role of a sizeable set of spatial variables in intersection crash occurrence. In addition to commonly considered traffic and geometric variables, examined spatial factors include local influences of weather, sun glare, proximity to drinking establishments, and proximity to schools. The results indicate that inclusion of these factors results in significant improvement in model explanatory power, and the results also generally agree with expectation. The research illuminates the importance of spatial variables in safety research and also the negative consequences of their omissions.

Property damage crash equivalency factors to solve crash frequency-severity dilemma: Case study on South Korean rural roads

Safety interventions (e.g., median barriers, photo enforcement) and road features (e.g., median type and width) can influence crash severity, crash frequency, or both. Both dimensions—crash frequency and crash severity—are needed to obtain a full accounting of road safety. Extensive literature and common sense both dictate that crashes are not created equal, with fatalities costing society more than 1,000 times the cost of property damage crashes on average. Despite this glaring disparity, the profession has not unanimously embraced or successfully defended a nonarbitrary severity weighting approach for analyzing safety data and conducting safety analyses. It is argued here that the two dimensions (frequency and severity) are made available by intelligently and reliably weighting crash frequencies and converting all crashes to property-damage-only crash equivalents (PDOEs) by using comprehensive societal unit crash costs. This approach is analogous to calculating axle load equivalents in the prediction of pavement damage: for instance, a 40,000-lb truck causes 4,025 times more stress than does a 4,000-lb car and so simply counting axles is not sufficient. Calculating PDOEs using unit crash costs is the most defensible and nonarbitrary weighting scheme, allows for the simple incorporation of severity and frequency, and leads to crash models that are sensitive to factors that affect crash severity. Moreover, using PDOEs diminishes the errors introduced by underreporting of less severe crashes—an added benefit of the PDOE analysis approach. The method is illustrated with rural road segment data from South Korea (which in practice would develop PDOEs with Korean crash cost data).

The significance of endogeneity problems in crash models: An examination of left-turn lanes in intersection crash models

Crash prediction models are used for a variety of purposes including forecasting the expected future performance of various transportation system segments with similar traits. The influence of intersection features on safety have been examined extensively because intersections experience a relatively large proportion of motor vehicle conflicts and crashes compared to other segments in the transportation system. The effects of left-turn lanes at intersections in particular have seen mixed results in the literature. Some researchers have found that left-turn lanes are beneficial to safety while others have reported detrimental effects on safety. This inconsistency is not surprising given that the installation of left-turn lanes is often endogenous, that is, influenced by crash counts and/or traffic volumes. Endogeneity creates problems in econometric and statistical models and is likely to account for the inconsistencies reported in the literature. This paper reports on a limited-information maximum likelihood (LIML) estimation approach to compensate for endogeneity between left-turn lane presence and angle crashes. The effects of endogeneity are mitigated using the approach, revealing the unbiased effect of left-turn lanes on crash frequency for a dataset of Georgia intersections. The research shows that without accounting for endogeneity, left-turn lanes ‘appear’ to contribute to crashes; however, when endogeneity is accounted for in the model, left-turn lanes reduce angle crash frequencies as expected by engineering judgment. Other endogenous variables may lurk in crash models as well, suggesting that the method may be used to correct simultaneity problems with other variables and in other transportation modeling contexts.

On the nature of over-dispersion in motor vehicle crash prediction models

Statistical modeling of traffic crashes has been of interest to researchers for decades. Over the most recent decade many crash models have accounted for extra-variation in crash counts—variation over and above that accounted for by the Poisson density. The extra-variation – or dispersion – is theorized to capture unaccounted for variation in crashes across sites. The majority of studies have assumed fixed dispersion parameters in over-dispersed crash models—tantamount to assuming that unaccounted for variation is proportional to the expected crash count. Miaou and Lord [Miaou, S.P., Lord, D., 2003. Modeling traffic crash-flow relationships for intersections: dispersion parameter, functional form, and Bayes versus empirical Bayes methods. Transport. Res. Rec. 1840, 31–40] challenged the fixed dispersion parameter assumption, and examined various dispersion parameter relationships when modeling urban signalized intersection accidents in Toronto. They suggested that further work is needed to determine the appropriateness of the findings for rural as well as other intersection types, to corroborate their findings, and to explore alternative dispersion functions. This study builds upon the work of Miaou and Lord, with exploration of additional dispersion functions, the use of an independent data set, and presents an opportunity to corroborate their findings. Data from Georgia are used in this study. A Bayesian modeling approach with non-informative priors is adopted, using sampling-based estimation via Markov Chain Monte Carlo (MCMC) and the Gibbs sampler. A total of eight model specifications were developed; four of them employed traffic flows as explanatory factors in mean structure while the remainder of them included geometric factors in addition to major and minor road traffic flows. The models were compared and contrasted using the significance of coefficients, standard deviance, chi-square goodness-of-fit, and deviance information criteria (DIC) statistics. The findings indicate that the modeling of the dispersion parameter, which essentially explains the extra-variance structure, depends greatly on how the mean structure is modeled. In the presence of a well-defined mean function, the extra-variance structure generally becomes insignificant, i.e. the variance structure is a simple function of the mean. It appears that extra-variation is a function of covariates when the mean structure (expected crash count) is poorly specified and suffers from omitted variables. In contrast, when sufficient explanatory variables are used to model the mean (expected crash count), extra-Poisson variation is not significantly related to these variables. If these results are generalizable, they suggest that model specification may be improved by testing extra-variation functions for significance. They also suggest that known influences of expected crash counts are likely to be different than factors that might help to explain unaccounted for variation in crashes across sites

New criteria for evaluating methods of identifying hot spots

Identification of hot spots, also known as the sites with promise, black spots, accident-prone locations, or priority investigation locations, is an important and routine activity for improving the overall safety of roadway networks. Extensive literature focuses on methods for hot spot identification (HSID). A subset of this considerable literature is dedicated to conducting performance assessments of various HSID methods. A central issue in comparing HSID methods is the development and selection of quantitative and qualitative performance measures or criteria. The authors contend that currently employed HSID assessment criteria—namely false positives and false negatives—are necessary but not sufficient, and additional criteria are needed to exploit the ordinal nature of site ranking data. With the intent to equip road safety professionals and researchers with more useful tools to compare the performances of various HSID methods and to improve the level of HSID assessments, this paper proposes four quantitative HSID evaluation tests that are, to the authors’ knowledge, new and unique. These tests evaluate different aspects of HSID method performance, including reliability of results, ranking consistency, and false identification consistency and reliability. It is intended that road safety professionals apply these different evaluation tests in addition to existing tests to compare the performances of various HSID methods, and then select the most appropriate HSID method to screen road networks to identify sites that require further analysis. This work demonstrates four new criteria using 3 years of Arizona road section accident data and four commonly applied HSID methods [accident frequency ranking, accident rate ranking, accident reduction potential, and empirical Bayes (EB)]. The EB HSID method reveals itself as the superior method in most of the evaluation tests. In contrast, identifying hot spots using accident rate rankings performs the least well among the tests. The accident frequency and accident reduction potential methods perform similarly, with slight differences explained. The authors believe that the four new evaluation tests offer insight into HSID performance heretofore unavailable to analysts and researchers.

Incorporating weather into regionwide safety planning prediction models

Predicting safety on roadways is standard practice for road safety professionals and has a corresponding extensive literature. The majority of safety prediction models are estimated using roadway segment and intersection (microscale) data, while more recently efforts have been undertaken to predict safety at the planning level (macroscale). Safety prediction models typically include roadway, operations, and exposure variables—factors known to affect safety in fundamental ways. Environmental variables, in particular variables attempting to capture the effect of rain on road safety, are difficult to obtain and have rarely been considered. In the few cases weather variables have been included, historical averages rather than actual weather conditions during which crashes are observed have been used. Without the inclusion of weather related variables researchers have had difficulty explaining regional differences in the safety performance of various entities (e.g. intersections, road segments, highways, etc.) As part of the NCHRP 8-44 research effort, researchers developed PLANSAFE, or planning level safety prediction models. These models make use of socio-economic, demographic, and roadway variables for predicting planning level safety. Accounting for regional differences - similar to the experience for microscale safety models - has been problematic during the development of planning level safety prediction models. More specifically, without weather related variables there is an insufficient set of variables for explaining safety differences across regions and states. Furthermore, omitted variable bias resulting from excluding these important variables may adversely impact the coefficients of included variables, thus contributing to difficulty in model interpretation and accuracy. This paper summarizes the results of an effort to include weather related variables, particularly various measures of rainfall, into accident frequency prediction and the prediction of the frequency of fatal and/or injury degree of severity crash models. The purpose of the study was to determine whether these variables do in fact improve overall goodness of fit of the models, whether these variables may explain some or all of observed regional differences, and identifying the estimated effects of rainfall on safety. The models are based on Traffic Analysis Zone level datasets from Michigan, and Pima and Maricopa Counties in Arizona. Numerous rain-related variables were found to be statistically significant, selected rain related variables improved the overall goodness of fit, and inclusion of these variables reduced the portion of the model explained by the constant in the base models without weather variables. Rain tends to diminish safety, as expected, in fairly complex ways, depending on rain frequency and intensity.

A simultaneous equations model of crash frequency by collision type for rural intersections

Safety at roadway intersections is of significant interest to transportation professionals due to the large number of intersections in transportation networks, the complexity of traffic movements at these locations that leads to large numbers of conflicts, and the wide variety of geometric and operational features that define them. A variety of collision types including head-on, sideswipe, rear-end, and angle crashes occur at intersections. While intersection crash totals may not reveal a site deficiency, over exposure of a specific crash type may reveal otherwise undetected deficiencies. Thus, there is a need to be able to model the expected frequency of crashes by collision type at intersections to enable the detection of problems and the implementation of effective design strategies and countermeasures. Statistically, it is important to consider modeling collision type frequencies simultaneously to account for the possibility of common unobserved factors affecting crash frequencies across crash types. In this paper, a simultaneous equations model of crash frequencies by collision type is developed and presented using crash data for rural intersections in Georgia. The model estimation results support the notion of the presence of significant common unobserved factors across crash types, although the impact of these factors on parameter estimates is found to be rather modest.

Examination of relationships between urban form, household activities, and time allocation in the Atlanta Metropolitan Region

A substantial body of research is focused on understanding the relationships between socio-demographics, land-use characteristics, and mode specific attributes on travel mode choice and time-use patterns. Residential and commercial densities, inter-mixing of land uses, and route directness in conjunction with transportation performance characteristics interact to influence accessibility to destinations as well as time spent traveling and engaging in activities. This study uniquely examines the activity durations undertaken for out-of-home subsistence; maintenance, and discretionary activities. Also examined are total tour durations (summing all activity categories within a tour). Cross-sectional activities are obtained from household activity travel survey data from the Atlanta Metropolitan Region. Time durations allocated to weekdays and weekends are compared. The censoring and endogeneity between activity categories and within individuals are captured using multiple equations Tobit models. The analysis and modeling reveal that land-use characteristics such as net residential density and the number of commercial parcels within a kilometer of a residence are associated with differences in weekday and weekend time-use allocations. Household type and structure are significant predictors across the three activity categories, but not for overall travel times. Tour characteristics such as time-of-day and primary travel mode of the tours also affect traveler's out-of-home activity-tour time-use patterns.

Poisson, Poisson-gamma and zero-inflated regression models of motor vehicle crashes: balancing statistical fit and theory

There has been considerable research conducted over the last 20 years focused on predicting motor vehicle crashes on transportation facilities. The range of statistical models commonly applied includes binomial, Poisson, Poisson-gamma (or negative binomial), zero-inflated Poisson and negative binomial models (ZIP and ZINB), and multinomial probability models. Given the range of possible modeling approaches and the host of assumptions with each modeling approach, making an intelligent choice for modeling motor vehicle crash data is difficult. There is little discussion in the literature comparing different statistical modeling approaches, identifying which statistical models are most appropriate for modeling crash data, and providing a strong justification from basic crash principles. In the recent literature, it has been suggested that the motor vehicle crash process can successfully be modeled by assuming a dual-state data-generating process, which implies that entities (e.g., intersections, road segments, pedestrian crossings, etc.) exist in one of two states—perfectly safe and unsafe. As a result, the ZIP and ZINB are two models that have been applied to account for the preponderance of “excess” zeros frequently observed in crash count data. The objective of this study is to provide defensible guidance on how to appropriate model crash data. We first examine the motor vehicle crash process using theoretical principles and a basic understanding of the crash process. It is shown that the fundamental crash process follows a Bernoulli trial with unequal probability of independent events, also known as Poisson trials. We examine the evolution of statistical models as they apply to the motor vehicle crash process, and indicate how well they statistically approximate the crash process. We also present the theory behind dual-state process count models, and note why they have become popular for modeling crash data. A simulation experiment is then conducted to demonstrate how crash data give rise to “excess” zeros frequently observed in crash data. It is shown that the Poisson and other mixed probabilistic structures are approximations assumed for modeling the motor vehicle crash process. Furthermore, it is demonstrated that under certain (fairly common) circumstances excess zeros are observed—and that these circumstances arise from low exposure and/or inappropriate selection of time/space scales and not an underlying dual state process. In conclusion, carefully selecting the time/space scales for analysis, including an improved set of explanatory variables and/or unobserved heterogeneity effects in count regression models, or applying small-area statistical methods (observations with low exposure) represent the most defensible modeling approaches for datasets with a preponderance of zeros

Probability and statistics

Now in its sixth edition, the Traffic Engineering Handbook continues to be a must have publication in the transportation industry, as it has been for the past 60 years. The new edition provides updated information for people entering the practice and for those already practicing. The handbook is a convenient desk reference, as well as an all in one source of principles and proven techniques in traffic engineering. Most chapters are presented in a new format, which divides the chapters into four areas-basics, current practice, emerging trends and information sources. Chapter topics include road users, vehicle characteristics, statistics, planning for operations, communications, safety, regulations, traffic calming, access management, geometrics, signs and markings, signals, parking, traffic demand, maintenance and studies. In addition, as the focus in transportation has shifted from project based to operations based, two new chapters have been added-"Planning for Operations" and "Managing Traffic Demand to Address Congestion: Providing Travelers with Choices." The Traffic Engineering Handbook continues to be one of the primary reference sources for study to become a certified Professional Traffic Operations Engineer™. Chapters are authored by notable and experienced authors, and reviewed and edited by a distinguished panel of traffic engineering experts.

Accident prediction model for railway-highway interfaces

Considerable past research has explored relationships between vehicle accidents and geometric design and operation of road sections, but relatively little research has examined factors that contribute to accidents at railway-highway crossings. Between 1998 and 2002 in Korea, about 95% of railway accidents occurred at highway-rail grade crossings, resulting in 402 accidents, of which about 20% resulted in fatalities. These statistics suggest that efforts to reduce crashes at these locations may significantly reduce crash costs. The objective of this paper is to examine factors associated with railroad crossing crashes. Various statistical models are used to examine the relationships between crossing accidents and features of crossings. The paper also compares accident models developed in the United States and the safety effects of crossing elements obtained using Korea data. Crashes were observed to increase with total traffic volume and average daily train volumes. The proximity of crossings to commercial areas and the distance of the train detector from crossings are associated with larger numbers of accidents, as is the time duration between the activation of warning signals and gates. The unique contributions of the paper are the application of the gamma probability model to deal with underdispersion and the insights obtained regarding railroad crossing related vehicle crashes. Considerable past research has explored relationships between vehicle accidents and geometric design and operation of road sections, but relatively little research has examined factors that contribute to accidents at railway-highway crossings. Between 1998 and 2002 in Korea, about 95% of railway accidents occurred at highway-rail grade crossings, resulting in 402 accidents, of which about 20% resulted in fatalities. These statistics suggest that efforts to reduce crashes at these locations may significantly reduce crash costs. The objective of this paper is to examine factors associated with railroad crossing crashes. Various statistical models are used to examine the relationships between crossing accidents and features of crossings. The paper also compares accident models developed in the United States and the safety effects of crossing elements obtained using Korea data. Crashes were observed to increase with total traffic volume and average daily train volumes. The proximity of crossings to commercial areas and the distance of the train detector from crossings are associated with larger numbers of accidents, as is the time duration between the activation of warning signals and gates. The unique contributions of the paper are the application of the gamma probability model to deal with underdispersion and the insights obtained regarding railroad crossing related vehicle crashes.

Forecasting crashes at the planning level: Simultaneous negative binomial crash model applied in Tucson, Arizona

At least two important transportation planning activities rely on planning-level crash prediction models. One is motivated by the Transportation Equity Act for the 21st Century, which requires departments of transportation and metropolitan planning organizations to consider safety explicitly in the transportation planning process. The second could arise from a need for state agencies to establish incentive programs to reduce injuries and save lives. Both applications require a forecast of safety for a future period. Planning-level crash prediction models for the Tucson, Arizona, metropolitan region are presented to demonstrate the feasibility of such models. Data were separated into fatal, injury, and property-damage crashes. To accommodate overdispersion in the data, negative binomial regression models were applied. To accommodate the simultaneity of fatality and injury crash outcomes, simultaneous estimation of the models was conducted. All models produce crash forecasts at the traffic analysis zone level. Statistically significant (p-values < 0.05) and theoretically meaningful variables for the fatal crash model included population density, persons 17 years old or younger as a percentage of the total population, and intersection density. Significant variables for the injury and property-damage crash models were population density, number of employees, intersections density, percentage of miles of principal arterial, percentage of miles of minor arterials, and percentage of miles of urban collectors. Among several conclusions it is suggested that planning-level safety models are feasible and may play a role in future planning activities. However, caution must be exercised with such models.

Differences in the performance of safety performance functions estimated for total crash count and for crash count by crash type

In recent years the development and use of crash prediction models for roadway safety analyses have received substantial attention. These models, also known as safety performance functions (SPFs), relate the expected crash frequency of roadway elements (intersections, road segments, on-ramps) to traffic volumes and other geometric and operational characteristics. A commonly practiced approach for applying intersection SPFs is to assume that crash types occur in fixed proportions (e.g., rear-end crashes make up 20% of crashes, angle crashes 35%, and so forth) and then apply these fixed proportions to crash totals to estimate crash frequencies by type. As demonstrated in this paper, such a practice makes questionable assumptions and results in considerable error in estimating crash proportions. Through the use of rudimentary SPFs based solely on the annual average daily traffic (AADT) of major and minor roads, the homogeneity-in-proportions assumption is shown not to hold across AADT, because crash proportions vary as a function of both major and minor road AADT. For example, with minor road AADT of 400 vehicles per day, the proportion of intersecting-direction crashes decreases from about 50% with 2,000 major road AADT to about 15% with 82,000 AADT. Same-direction crashes increase from about 15% to 55% for the same comparison. The homogeneity-in-proportions assumption should be abandoned, and crash type models should be used to predict crash frequency by crash type. SPFs that use additional geometric variables would only exacerbate the problem quantified here. Comparison of models for different crash types using additional geometric variables remains the subject of future research.

Use of expert panels in highway safety: A critique

Expert panels have been used extensively in the development of the "Highway Safety Manual" to extract research information from highway safety experts. While the panels have been used to recommend agendas for new and continuing research, their primary role has been to develop accident modification factors—quantitative relationships between highway safety and various highway safety treatments. Because the expert panels derive quantitative information in a “qualitative” environment and because their findings can have significant impacts on highway safety investment decisions, the expert panel process should be described and critiqued. This paper is the first known written description and critique of the expert panel process and is intended to serve professionals wishing to conduct such panels.

Expected safety performance of rural signalized intersections in South Korea

Understanding the expected safety performance of rural signalized intersections is critical for (a) identifying high-risk sites where the observed safety performance is substantially worse than the expected safety performance, (b) understanding influential factors associated with crashes, and (c) predicting the future performance of sites and helping plan safety-enhancing activities. These three critical activities are routinely conducted for safety management and planning purposes in jurisdictions throughout the United States and around the world. This paper aims to develop baseline expected safety performance functions of rural signalized intersections in South Korea, which to date have not yet been established or reported in the literature. Data are examined from numerous locations within South Korea for both three-legged and four-legged configurations. The safety effects of a host of operational and geometric variables on the safety performance of these sites are also examined. In addition, supplementary tables and graphs are developed for comparing the baseline safety performance of sites with various geometric and operational features. These graphs identify how various factors are associated with safety. The expected safety prediction tables offer advantages over regression prediction equations by allowing the safety manager to isolate specific features of the intersections and examine their impact on expected safety. The examination of the expected safety performance tables through illustrated examples highlights the need to correct for regression-to-the-mean effects, emphasizes the negative impacts of multicollinearity, shows why multivariate models do not translate well to accident modification factors, and illuminates the need to examine road safety carefully and methodically. Caveats are provided on the use of the safety performance prediction graphs developed in this paper.