Winter 2001
Eitan Naveh
Alfred Marcus
Airplane crashes and accidents at nuclear power plants are thoroughly analyzed to determine how they might be prevented. The average person, however, is much more likely to be involved in a car crash, and yet these incidents and their causes do not receive the same scrutiny, say a pair of University of Minnesota researchers.
Principal investigator Eitan Naveh, an instructor and researcher formerly with the Center for the Development of Technological Leadership, and Professor Alfred Marcus of the Carlson School of Management set out to examine and assess how current crash data are collected in Minnesota. Their goal was to discover what more could be learned from already-existing data, what data is missing, and how data might be collected and better used in order to improve road safety.
Naveh and Marcus began by analyzing crash data from the Minnesota Department of Transportation's Minnesota Motor Vehicle Crash Facts, an annual publication of data from accident reports filed by law enforcement officers for any crash resulting in injuries or in $1,000 or more in damage. In 1999, there were nearly 93,000 crashes in Minnesota, with approximately 45,000 injuries and 650 fatalities.
In addition to examining the factors involved in crashes, the researchers analyzed the Minnesota driver's instruction manual to see if what they discovered about the causes of crashes correlated to information in the manual. What they found was that less than one-third of the driving manual is devoted to the main causes of crashes as indicated by the data collected and analyzed. "We feel this [crash] data should be used to initiate learning processes," Naveh says. An overall tightening of the system with improvements in data collection and analysis, and closer ties between the collection and analysis of the data and its use in such programs as driver education, may be able to reduce traffic fatalities and injuries.
The researchers also found that it is not always clear who the customers for the data are. They recommend better communication and cooperation among the different organizations that collect and use data, and a sharp and clear definition of customers and users. Additionally, crash data should be communicated in waysthrough driver's education, news reports, public service announcements, and the likethat could help ordinary citizens improve their driving ability and avoid crashes.
The crash facts publication focuses on the factors that cause crashes. In the second part of their research, Naveh and Marcus used data on crashes in Minnesota to find the answers to a slightly different pair of questions. They asked, given that a crash has occurred, what factors are most important in determining whether injuries or fatalities result? To illustrate the difference, consider the examples of icy roads and not wearing seat belts. Icy roads make it more likely that a crash will occur, but will not cause the crash to be more severe, because people tend to drive slower on ice. By contrast, not wearing a seat belt will not increase the likelihood of a crash occurring, but will substantially increase the odds of injury or fatality if it does.
The researchers developed two modelsone for crashes with fatalities, one for crashes with injuriesand examined the conditions faced and decisions made by drivers. In both models, they found the most significant condition was rural roads, and the most significant driver decision was not using a seat belt. Overall, Naveh and Marcus found that the crash data explained the likelihood of fatalities or injuries fairly well. Nonetheless, the data collection process could be improved, they say, perhaps by implementing consistent standards or using technology to improve the accuracy of on-site data collection and to assist in identifying crash locations.
"If we can study and understand what sort of decisions and reactions are needed in certain conditions, we can improve safety," Naveh said.