Controlling for Self-reported Exposure in Traffic Accident Prediction Studies

For many decades, traffic safety researchers have been trying to predict collision involvement using individual differences (e.g., Clarke and Robertson, 2005; Janke, 1993; Owsley and McGwin, 1999; Tillman and Hobbs, 1949; see the reviews by Adams, 1970). In this work, exposure to risk of accident has most often not been controlled for beyond the basic method of calculating crashes per year. Although exposure has often been discussed and researched in various ways (Brown, 1982; Chipman, 1982; Chipman, MacGregor, Smiley and Lee-Gosselin, 1992; Janke, 1991; af Wåhlberg, 2009), actually controlling for it in accident prediction studies with more exact methods, such as mileage, has been relatively uncommon (af Wåhlberg, 2009).