Naveen Eluru

A mixed generalized ordered response model for examining pedestrian and bicyclist injury severity level in traffic crashes.

This paper proposes an econometric structure for injury severity analysis at the level of individual accidents that recognizes the ordinal nature of the categories in which injury severity are recorded, while also allowing flexibility in capturing the effects of explanatory variables on each ordinal category and allowing heterogeneity in the effects of contributing factors due to the moderating influence of unobserved factors. The model developed here, referred to as the mixed generalized ordered response logit (MGORL) model, generalizes the standard ordered response models used in the extant literature for injury severity analysis. To our knowledge, this is the first such formulation to be proposed and applied in the econometric literature in general, and in the safety analysis literature in particular. The MGORL model is applied to examine non-motorist injury severity in accidents in the USA, using the 2004 General Estimates System (GES) database. The empirical findings emphasize the inconsistent results obtained from the standard ordered response model. An important policy result from our analysis is that the general pattern and relative magnitude of elasticity effects of injury severity determinants are similar for pedestrians and bicyclists. The analysis also suggests that the most important variables influencing non-motorist injury severity are the age of the individual (the elderly are more injury-prone), the speed limit on the roadway (higher speed limits lead to higher injury severity levels), location of crashes (those at signalized intersections are less severe than those elsewhere), and time-of-day (darker periods lead to higher injury severity).

A latent class modeling approach for identifying vehicle driver injury severity factors at highway-railway crossings

In this paper, we aim to identify the different factors that influence injury severity of highway vehicle occupants, in particular drivers, involved in a vehicle-train collision at highway-railway grade crossings. The commonly used approach to modeling vehicle occupant injury severity is the traditional ordered response model that assumes the effect of various exogenous factors on injury severity to be constant across all accidents. The current research effort attempts to address this issue by applying an innovative latent segmentation based ordered logit model to evaluate the effects of various factors on the injury severity of vehicle drivers. In this model, the highway-railway crossings are assigned probabilistically to different segments based on their attributes with a separate injury severity component for each segment. The validity and strength of the formulated collision consequence model is tested using the US Federal Railroad Administration database which includes inventory data of all the railroad crossings in the US and collision data at these highway railway crossings from 1997 to 2006. The model estimation results clearly highlight the existence of risk segmentation within the affected grade crossing population by the presence of active warning devices, presence of permanent structure near the crossing and roadway type. The key factors influencing injury severity include driver age, time of the accident, presence of snow and/or rain, vehicle role in the crash and motorist action prior to the crash.

Examining the influence of aggressive driving behavior on driver injury severity in traffic crashes

In this paper, we capture the moderating effect of aggressive driving behavior while assessing the influence of a comprehensive set of variables on injury severity. In doing so, we are able to account for the indirect effects of variables on injury severity through their influence on aggressive driving behavior, as well as the direct effect of variables on injury severity. The methodology used in the paper to accommodate the moderating effect of aggressive driving behavior takes the form of two models--one for aggressive driving and another for injury severity. These are appropriately linked to obtain the indirect and direct effects of variables. The data for estimation is obtained from the National Motor Vehicle Crash Causation Study (NMVCCS). From an empirical standpoint, we consider a fine age categorization until 20 years of age when examining age effects on aggressive driving behavior and injury severity. There are several important results from the empirical analysis undertaken in the current paper based on post-crash data collection on aggressive behavior participation just prior to the crash and injury severity sustained in a crash. Young drivers (especially novice drivers between 16 and 17 years of age), drivers who are not wearing seat belt, under the influence of alcohol, not having a valid license, and driving a pick-up are found to be most likely to behave aggressively. Situational, vehicle, and roadway factors such as young drivers traveling with young passengers, young drivers driving an SUV or a pick-up truck, driving during the morning rush hour, and driving on roads with high speed limits are also found to trigger aggressive driving behavior. In terms of vehicle occupants, the safest situation from a driver injury standpoint is when there are two or more passengers in the vehicle, at least one of whom is above the age of 20 years. These and many other results are discussed, along with implications of the result for graduated driving licensing (GDL) programs.

Evaluating alternate discrete outcome frameworks for modeling crash injury severity

This paper focuses on the relevance of alternate discrete outcome frameworks for modeling driver injury severity. The study empirically compares the ordered response and unordered response models in the context of driver injury severity in traffic crashes. The alternative modeling approaches considered for the comparison exercise include: for the ordered response framework-ordered logit (OL), generalized ordered logit (GOL), mixed generalized ordered logit (MGOL) and for the unordered response framework-multinomial logit (MNL), nested logit (NL), ordered generalized extreme value logit (OGEV) and mixed multinomial logit (MMNL) model. A host of comparison metrics are computed to evaluate the performance of these alternative models. The study provides a comprehensive comparison exercise of the performance of ordered and unordered response models for examining the impact of exogenous factors on driver injury severity. The research also explores the effect of potential underreporting on alternative frameworks by artificially creating an underreported data sample from the driver injury severity sample. The empirical analysis is based on the 2010 General Estimates System (GES) data base-a nationally representative sample of road crashes collected and compiled from about 60 jurisdictions across the United States. The performance of the alternative frameworks are examined in the context of model estimation and validation (at the aggregate and disaggregate level). Further, the performance of the model frameworks in the presence of underreporting is explored, with and without corrections to the estimates. The results from these extensive analyses point toward the emergence of the GOL framework (MGOL) as a strong competitor to the MMNL model in modeling driver injury severity.

Integration of Activity-Based Modeling and Dynamic Traffic Assignment

The traditional trip-based approach to transportation modeling has been used for the past 30 years. Because of limitations of traditional planning for short-term policy analysis, researchers have explored alternative paradigms for incorporating more behavioral realism in planning methodologies. On the demand side, activity-based approaches have evolved as an alternative to traditional trip-based transportation demand forecasting. On the supply side, dynamic traffic assignment models have been developed as an alternative to static assignment procedures. Much of the research effort in activity-based approaches (the demand side) and dynamic traffic assignment techniques (the supply side) has been undertaken relatively independently. To maximize benefits from these advanced methodologies, it is essential to combine them through a unified framework. The objective of this paper is to develop a conceptual framework and explore practical integration issues for combining the two streams of research. Technical, computational, and practical issues involved in this demand–supply integration problem are discussed. The framework is general, but specific technical details related to the integration are explored by using CEMDAP for activity-based modeling and VISTA for dynamic traffic assignment modeling. Solution convergence properties of the integrated system, specifically examining different criteria for convergence, different methods of accommodating time of day, and the influence of step size on convergence are studied. The integrated system developed is empirically applied to two sample networks selected from the Dallas–Fort Worth system in Texas.

Joint Model of Choice of Residential Neighborhood and Bicycle Ownership: Accounting for Self-Selection and Unobserved Heterogeneity

This paper presents a joint model of residential neighborhood type choice and bicycle ownership. The objective is to isolate the true causal effects of neighborhood attributes on household bicycle ownership from a spurious association because of residential self-selection effects. The joint model accounts for residential self-selection because of both observed sociodemographic characteristics and unobserved preferences. In addition, the model allows differential residential self-selection effects across different sociodemographic segments. The model was estimated by using a sample of more than 5,000 households from the San Francisco, California, Bay Area. Furthermore, a policy simulation analysis was carried out to estimate the impacts of neighborhood characteristics and sociodemographics on bicycle ownership. The model results show a substantial presence of residential self-selection effects because of observed sociodemographics, such as the number of children, dwelling type, and house ownership. It is shown for the first time in the self-selection literature that ignoring such observed self-selection effects may not always lead to overestimation of the impact of neighborhood attributes on travel-related choices, such as bicycle ownership. In the current context, ignoring self-selection because of sociodemographic attributes resulted in an underestimation of the impact of neighborhood attributes on bicycle ownership. In the context of unobserved factors, no significant self-selection effects were found. However, it is recommended that such effects as well as the heterogeneity in such effects be tested for before it is concluded that there are no unobserved factors contributing to residential self-selection.

Modeling Interdependence in Household Residence and Workplace Choices

Models of residential and workplace location choice prevalent in the literature often assume that one choice dimension is exogenous to the other. But a broad and uniform assumption that one choice dimension is exogenous and influences the other may be too strong to use as the foundation for current behavioral research or applied policy analysis. This paper examines the interdependence of residence and workplace choices and develops a novel approach to modeling these choice dependencies. Two problems related to such joint modeling efforts are addressed. First, through a latent market segment modeling approach, the paper offers a methodology for accommodating different sequential decision-making processes that may be present in the population–for example, residential location may be chosen first and may influence workplace location for one segment and vice versa. Second, the modeling approach offers a means of overcoming the exploding choice set problem when attempting to model multidimensional choice phenomena. The overall aim of the work is to model the structure of the interdependency between the choices that a household makes about residence location and the workplace choices of the workers in the household in the context of an integrated activity location and travel forecasting framework. This paper presents a joint model of residence location and workplace using activity-based travel survey data collected in the Puget Sound region of Washington state in 1999, with novel adaptation of recent methods for incorporating latent market segmentation within discrete choice models.

Who Are Bicyclists? Why and How Much Are They Bicycling?

The factors influencing the decision to bicycle are explored and unraveled to inform the development of appropriate and effective strategies to increase bicycling and promote the health of individuals and of the environment. The data used in the analysis were drawn from a survey of Texas bicyclists, and the study includes a comprehensive explanatory analysis of bicyclists and their bicycling habits. Various econometric models are used to evaluate the determinants of bicyclists’ perception for safety and quality issues and the frequency of bicycling for commute and noncommute purposes. The results of the study indicate that the perceptions of the quality of bicycle facilities and safety from traffic crashes show significant variation depending on bicyclists’ demographic and work characteristics and bicycle amenities and facilities on the commute route and at the workplace. Bicyclist demographics (gender, age, education level, commute distance), household demographics (number of automobiles, number of bicycles, number of children), residential location and season, bicycle amenities at work (bicycle racks, showers), bicyclist perceptions of the overall quality of bicycle facilities, and bicycle-use characteristics affect commute and noncommute bicycling frequency. These study results can assist in the development of informed policies to increase commute and noncommute bicycling, and the results highlight the ongoing need for detailed surveys to understand bicycling behavior.

Modeling Injury Severity of Multiple Occupants of Vehicles: Copula-Based Multivariate Approach

Research to date on crash injury severity has focused on the driver of the vehicle or the most severely injured occupant. Though useful, these studies have not provided injury profiles of all occupants in crash-involved vehicles. This lack of a comprehensive picture has limited the ability to devise measures that enhance the safety and reduce the severity of the injury sustained by all vehicular occupants. Moreover, such studies ignore the possible presence of correlated, unobserved factors that may simultaneously affect the injury severity levels of multiple occupants. This paper aims to fill the gap by presenting a simultaneous model of injury severity to apply to crashes that involve any number of occupants. A copula-based methodology, which could be used to estimate such complex model systems, was applied to a data set of crashes drawn from the 2007 General Estimates System in the United States. The model estimation results provide strong evidence of the presence of correlated unobserved factors that affect injury severity levels of vehicle occupants. The correlation exhibited heterogeneity across vehicle types, with a greater level of interoccupant dependency in heavy SUVs and pickup trucks. The study also sheds light on how numerous exogenous factors—including occupant, vehicle, and crash characteristics; environmental factors; and roadway attributes—affect the injury severity levels of occupants in different seat positions. The findings confirm that rear-seat passengers are less vulnerable to severe injuries than front-row passengers and point to the need to enhance vehicular design features that promote front-row occupant safety.

Macro-level pedestrian and bicycle crash analysis: Incorporating spatial spillover effects in dual state count models

This study attempts to explore the viability of dual-state models (i.e., zero-inflated and hurdle models) for traffic analysis zones (TAZs) based pedestrian and bicycle crash frequency analysis. Additionally, spatial spillover effects are explored in the models by employing exogenous variables from neighboring zones. The dual-state models such as zero-inflated negative binomial and hurdle negative binomial models (with and without spatial effects) are compared with the conventional single-state model (i.e., negative binomial). The model comparison for pedestrian and bicycle crashes revealed that the models that considered observed spatial effects perform better than the models that did not consider the observed spatial effects. Across the models with spatial spillover effects, the dual-state models especially zero-inflated negative binomial model offered better performance compared to single-state models. Moreover, the model results clearly highlighted the importance of various traffic, roadway, and sociodemographic characteristics of the TAZ as well as neighboring TAZs on pedestrian and bicycle crash frequency.