Identifying factors that predict teen driver crashes

Reducing crashes, in particular those that result in injury or fatality, is an ongoing struggle for agencies tasked with making our roads safer. The ability to predict these crashes would allow the agencies to develop interventions to change drivers' behavior and ultimately reduce the number of these crashes. Teen drivers are ideally suited for this type of intervention for several reasons. They are disproportionately over-represented in crashes -- teens account for only 4% of the driver population but account for 10% of crashes. Also, many teen crashes occur within the first year of being licensed and are the result of a lack of driving experience. Lastly, the learning curve is still steep at this point in teens' driving history which makes them more susceptible to interventions. In an attempt to predict these crashes, North Dakota driver licensing data and crash data were used to develop a sample of 20,392 teen drivers age 14 to 17. Within the first year after being licensed, these drivers sustained 317 crashes that resulted in an injury or death. The resulting logistic regression model identifies gender, traffic convictions, rural/urban, geography, and involvement in previous property-damage-only (PDO) crashes as markers that are significant in predicting these injury and fatal crashes.