DRIVER RECORDS AND CRASH PREDICTION

No method currently exists to predict with accuracy the crash involvement of individuals, even when crash records are augmented by other variables such as age and sex. As mentioned earlier, one factor in this low level of predictability is the huge variation in police accident reporting procedures. However, even with better measures of crash experience, the predictive accuracy for individuals would remain low. Some researchers have suggested that even the best predictive variables theoretically could explain no more than 10 to 15 percent of the variation in three-year crash experience because of the large size of the random component in individual crash likelihood. This is not to say that there are not identifiable groups of drivers with elevated crash risk. Problem drivers with very deviant records are several times more likely to crash than other drivers, and barring them from driving will eliminate some crashes. However, these people account for such a small segment of the problem that preventing them from driving can have little direct effect on overall crash totals. Thus, for actions such as license suspension or revocation to be effective in reducing highway crashes, they must serve as effective deterrents to other drivers. Whether they are effective as deterrents has yet to be determined.

Media Info

  • Features: Figures; References;
  • Pagination: 4 p.

Subject/Index Terms

Filing Info

  • Accession Number: 00395307
  • Record Type: Publication
  • Source Agency: National Highway Traffic Safety Administration
  • Report/Paper Numbers: Research Note 106, HS-037 889
  • Files: HSL, TRIS, USDOT
  • Created Date: Jun 30 1985 12:00AM