The Contribution of Driving Exposure in the Prediction of Traffic Accidents
The Contribution of Driving Exposure in the Prediction of Traffic Accidents
By Michael Gebers
Traffic safety research has shown that drivers differ in the amount of time they spend driving, the number of miles they drive, and the traffic conditions and situations in which they drive. For example, in a 1991 article published in Accident Analysis and Prevention, Janke noted that people driving low mileage tend to accumulate much of their mileage on congested city streets with two-way traffic and no restricted access and that high-mileage drivers typically accumulate most of those miles on freeways or other divided multilane highways with limited access. Such differences in exposure are seldom taken into account in studies on accident risk due to the difficulty of collecting accurate data on exposure variables.
In a Research and Development Branch (R&D) study using exposure data obtained from a random sample of licensed California drivers, Peck and Kuan (1983) reported an almost 14% increase in accident prediction after adding exposure variables to a regression equation containing driver record information alone. The results from this and other traffic safety studies indicate that exposure measures such as mileage are important correlates of accident risk that should not be overlooked in the development of accident risk models.
To further investigate the relationship between driving exposure and the risk of traffic accident involvement, R&D recently completed a study assessing the value of using exposure data in predicting traffic accident risk. Specifically, analyses were conducted to determine if the inclusion of driving exposure data collected in Hennessy's (1995) R&D study of experimental vision tests for license renewal applicants improved the prediction of total accidents beyond that achieved by using only prior accident and citation history available on driving records. The candidate exposure variables included miles driven, type of roadway (freeway versus city streets), driving time (night versus day), and traffic conditions (heavy versus light).
Prior research has found that the relationship between miles driven and traffic accidents is not linear but rather follows a negatively accelerating curve. Janke, for example, explained that the relationship between miles driven and traffic accidents is nonlinear, with smaller proportional increases in accident rate at higher levels of mileage. There are at least three possible explanations for this nonlinear relationship. First, it may be that drivers with higher mileage drive a greater number of these miles on relatively safe roads such as highways. Second, higher mileage drivers may be more skilled, possibly by virtue of their greater driving experience. Third, higher mileage drivers may adopt a relatively safer driving style, thereby reducing their risk of traffic accidents.
As illustrated in the figure below, the present report confirmed the nonlinear relationship between miles driven and traffic accident involvement. The figure shows that the accident risk curve dips for drivers reporting higher mileage.
The report presented several statistical models showing the relationship between driver record and exposure variables. The results indicate that the prediction of traffic accident involvement is enhanced by including both exposure and driving habit information. It was demonstrated that the prediction of traffic accidents nearly doubled after adding driving exposure/habit variables to the equation containing driver record histories.
The results of this study provide evidence that a driver's current level of driving exposure can influence their risk of accident involvement. For example, the combined driver record/survey equations indicate that driving 2 or fewer days per week was associated with a significantly lower risk of subsequent accident involvement. The results also indicate that drivers who avoid driving at night and driving in rainy or foggy conditions were less likely to be involved in subsequent accidents than were drivers who did not avoid such driving. This connection between lower accident involvement and limiting driving exposure could be exploited in driver education programs and in driver licensing. That is, more driver training and practice on limited access roads and during restricted hours of the day while learning to drive may prove beneficial in reducing accident involvements. Requiring more supervised driving time before licensure also limits exposure and would be expected to reduce accident risk. These and other strategies that reduce risk exposure while initial driving experience is being gained are part of California's graduated licensing program for drivers under 18 years of age.
That there may be additional benefits of using measures of exposure to assess accident risk was evidenced in the finding that subsequent accident risk decreased across the mileage categories with increasing numbers of prior citations. For example, it was found that among drivers with four traffic citations in the prior 2 years, the probability of subsequent accident involvement decreased as weekly miles driven increased.
The report suggests that the methodology and statistical modeling techniques used in the study would serve as a useful guide in the development and evaluation of a driving exposure and habits survey targeting a random sample of all licensed drivers. This survey is anticipated to be conducted within the next few years.
