CONTEXT. Implementation of graduated driver licensing programs is associated with reductions in crash rates of young drivers, but graduated driver licensing programs vary in their components. The impact of programs with different components is unknown.
OBJECTIVE. The purpose of this work was to determine which graduated driver licensing programs are associated with the greatest reductions in fatal motor vehicle crashes involving 16-year-old drivers.
METHODS. We conducted a retrospective study of all 16-year-old drivers involved in fatal crashes in the United States from 1994 through 2004 using data from the Fatality Analysis Reporting System and the US Census Bureau. We measured incidence rate ratios of fatal motor vehicle crashes involving 16-year-old drivers according to graduated driver licensing programs, adjusted for state and year.
RESULTS. Compared with state quarters with no graduated driver licensing program components, reductions of 16% to 21% in fatal crash involvement rates of 16-year-old drivers occurred with programs that included ≥3-month mandatory waiting period, nighttime driving restriction, and either ≥30 hours of supervised driving or passenger restriction. Reductions of 18% to 21% occurred in state quarters with programs that included ≥5 of the 7 components examined. Drivers aged 20 to 24 or 25 to 29 years did not experience significant reductions.
CONCLUSION. Comprehensive graduated driver licensing programs are associated with reductions of ∼20% in 16-year-old drivers’ fatal crash involvement rates. The greatest benefit seems to be associated with programs that include age requirements and ≥3 months of waiting before the intermediate stage, nighttime driving restriction, and either ≥30 hours of supervised driving or passenger restriction.
The Chen, Baker & Li (2006) article examining the effects of graduated driver licensing (GDL) on young driver fatalities is a valuable contribution to the adolescent injury prevention literature. However, it is important to avoid misinterpreting their findings when formulating teen driver policy.
This is a somewhat technical issue regarding researcher choices concerning details of how to conduct statistical analyses. In the present case, the matter is particularly important because the findings of single studies on young drivers are often quickly used by policy makers without attending to the larger body of literature on adolescent behavior in general, and young drivers in particular. The beneficial clarifying and filtering effect of research findings making their way to the general public through the professional research community is thus precluded.
The specific issue at hand is the authors' decision to use three months as the cut-point below which states were not considered to require a minimum duration learner permit (which allows only adult-supervised driving). The choice of three months is arbitrary. As a research decision, there is nothing wrong with that choice, but the findings should not be interpreted as suggesting that three months is sufficient to achieve the effect reported (although that is the clear, if unintended, implication). Nearly all states that fall into the category of "requiring at least three months supervised driving" in fact require six months or longer as the minimum learner period. Only one state that includes a minimum mandatory learner period requires as little as three months and only two others require less than six months; six months is by far the most common duration for a mandatory learner permit. Hence, the reported findings - with respect to learner permit periods - are more appropriately interpreted as resulting from learner periods of six months (or more) rather than three months.
The issue here is a case of the well-known and understood "grouping error," the resulting loss of information any time a continuum is treated as categorical or when many categories are collapsed into fewer categories. This is often necessary for analytic purposes, but information is inevitably lost in doing so. Researchers understand this and are cautious in their conclusions as a result, being careful not to attribute particular meaning to the precise cut-points and value ranges that have been chosen as a matter of necessity, rather than because of their inherent meaningfulness. In the present case, had the authors chosen 4 months, 5 months, or 6 months as their cut point, the results would have been essentially the same as were found using three months. Were any state policy makers to conclude that three months is sufficient to achieve the effects reported in this article, subsequently reducing a longer permit period to three months as a result, it would be a serious mistake, based on incorrect interpretation of the present findings.
Conflict of Interest:
None declared