Exploring American Drivers' Crash Histories: Transportation Data Acquisition and Analysis
Professor Kara Kockelman's graduate course set out to explore crash histories so that personal characteristics could be better linked to crash risks.
Traffic accidents are a leading cause of fatalities worldwide, accounting for 1.3 million deaths annually. While the number of fatalities is falling in many developed nations (thanks largely to new vehicle designs and roadway improvements), total traffic-related deaths are rising, as emerging economies grow and driving becomes more popular. In the U.S., motor vehicle crashes take more young lives than any other factor. In addition to these human costs, economic impacts are substantial, and experts estimate average crash costs at roughly $1,000 per American per year. Such figures offer important motivation for transportation engineers, planners, policymakers, and vehicle manufacturers to make transport safer.
In the U.S., police accident reports serve as the basis for local and nationwide crash databases. Only basic information on crash victims is provided, such as age, gender, seating position and seat belt use. When crash databases are linked to roadway network and vehicle features, researchers are able to estimate crash frequencies and severities as a function of roadway and vehicle design attributes. But there are no databases that capture the crash histories of those involved, so researchers have been unable to robustly link personal characteristics to crash risks.
Professor Kara Kockelman’s graduate course in the Acquisition and Analysis of Transportation Data (CE 392E) set out to explore such relationships by designing a survey to query respondents about their driving experiences and habits, annual miles traveled, traffic citation histories, crash histories (including those of family members and friends), vehicle ownership, safety-policy opinions, and demographics. After some review by experts in the field, the 130-question pilot survey was distributed to an Austin-based sample, resulting in an initial data set of 475 persons. Approximately 10 percent of these responses came from students in the course interview people directly, at various community events, at businesses, and while walking door-to-door. The remaining responses were collected through the survey website which was advertised to Austin neighborhood groups, university organizations, and other groups. The students learned first-hand the nuances of question structuring and what approaches can be most effective in recruiting respondents.
Preliminary data analysis using the Austin-based sample helped students identify strengths and weaknesses of the pilot survey. An improved survey with 100 questions was then marketed to a nation-wide 1000-respondent sample.
Initial data analysis suggests that the average American has experienced 2.1 crashes, with one of those having occurred in the last 10 years. Just 1 out of 200 crashes has resulted in a fatal injury, with another 4.4 percent resulting in severe injury, and 27.3 percent leading to minor injury. Models for crash counts suggest that younger and older respondents, those with more moving violations, those with higher household incomes, and those with two-door cars at home tend to be more frequently involved in crashes. Those who tend to drive more miles each year, sleep more hours each night, consume less alcohol, reside in larger households, have more minivans or trucks in their households, and live in more densely populated zip codes are less likely to have experienced crashes.
The work also examines respondent driving habits to deduce how personal characteristics influence crash exposure, speed choices, and higher risk of distracted driving (such as regular cellphone use, texting, and/or eating and drinking while driving).
For more, please visit the webpage of Kara Kockelman to view a pre-print of a technical paper and the survey instrument along with related safety articles.