Graduate Thesis Or Dissertation


Modeling Unobserved Heterogeneity and the Injury Severities of Truck Drivers in Run-Off-Road (ROR) Crashes: Econometric Methods and Applications Public Deposited

Downloadable Content

Download PDF


Attribute NameValues
  • Recent statistics regarding large truck crashes reveal that fatality rates of large trucks per 100 million vehicle miles traveled (VMT) and fatality rates per 1,000 registered vehicles are higher than those for passenger vehicles. These statistics underscore the need for greater efforts by safety professionals to help mitigate the impacts of these types of crashes on society and on the ground freight transportation industry as a whole. Of special interest are run-off-road (ROR) crashes (also referred to as roadway departure), which are crashes that occur due to a vehicle crossing an edge line or a center line of a roadway or/and leaving the designated lane. These types of crashes roughly constituted 54% of all traffic fatalities in the U.S. for the period between 2013 and 2015. There have been several efforts that have addressed large truck-involved crashes from varying perspectives (e.g., time of day, facility type, and single vs. multivehicle, work zone). However, there are several research gaps that need further attention, for example, better understanding of the relationship between contributing factors and driver injury severities due to ROR crashes involving large trucks. Specifically, those driver injury severities related to the impact of lighting conditions (dark vs. lighted) and land use (urban vs. rural). From a methodological modeling perspective accounting for unobserved heterogeneity has traditionally been ignored in regard to ROR driver injury severity analyses. To address these gaps, this dissertation aims to develop and estimate advanced econometric models that account for unobserved heterogeneity to better understand the contributing factors of driver injury severity of ROR crashes involving large trucks in the state of Oregon. This dissertation includes three manuscripts that investigate injury severity of large truck drivers involved in ROR crashes in the state of Oregon for the period 2007 to 2014. In the first manuscript, an ordered random parameter probit model was estimated to predict the likelihood of three injury severity categories using Oregon crash data: severe injury (fatal and incapacitating), minor injury (non-incapacitating and possible injury), and no injury while addressing the unobserved heterogeneity. The modeling framework presented in this manuscript offers a flexible methodology to analyze ROR crashes involving large trucks while accounting for unobserved heterogeneity. The second manuscript examines the impact of lighting conditions on injury severity of large truck drivers involved in ROR crashes. This was done by disaggregating crash data by lighting conditions into two datasets: one for the lighted condition and the other for the dark condition. Hence, two separate mixed logit models were developed to capture the contributing factors that affect injury severity in each lighting condition while accounting for unobserved heterogeneity. To validate the estimation results, series of likelihood ratio tests were conducted. Model separation tests along with estimation results indicate that lighting conditions need to be analyzed separately with 99.99% confidence. Lastly, an in-depth analysis was conducted in the third manuscript to examine the effect of land use setting (urban vs. rural) on injury severity of large truck drivers involved in ROR crashes. Again, disaggregating crash data was achieved to create two independent datasets: one pertaining to ROR crashes involving large trucks occurring on urban and the other for those occurring on rural areas. Instead of utilizing random parameter approach as a framework to account for unobserved heterogeneity, this manuscript utilizes two latent class ordered probit models to capture factors exclusively that contribute to each land use type. Once again, model separation tests along with estimation results reveal that there are distinctions in terms of contributing factors based on land use type. Therefore, ROR crashes involving large trucks need to be analyzed separately based on land use with 99.99% confidence. It is expected that estimating advanced econometric methods to identify contributing factors to injury severity of large truck drivers involved in ROR crashes while accounting for unobserved heterogeneity can be used as a basis to aid transportation safety engineers, trucking industry, transportation planners, and state agencies in implementing appropriate safety countermeasures to help mitigate ROR crashes. In terms of study implications, the estimation results of this dissertation have direct implications on safety policy. For instance, the finding of this dissertation reinforces the notion that the current policy regarding seatbelt usage should be amended in an attempt to increase seatbelt usage by penalizing drivers violating this policy to deterrent fines and forcing them to enroll in a required defensive driving course in the state of Oregon.
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Committee Member
Academic Affiliation
Rights Statement
Peer Reviewed



This work has no parents.

In Collection: