The relative risk of involvement in fatal crashes increases with increasing driver blood alcohol content and the risks increase more steeply for drivers under age 21 than for older drivers (Zador, Krawchuk and Voas, 2000). Previous studies have shown that raising the legal drinking age to 21 reduced drinking and alcohol related motor vehicle crash involvement among persons under age 21 (General Accounting Office, 1987; National Highway Traffic Safety Administration, 1999). Prior research also indicated that persons who grew up in states with a legal drinking age of 21 relative to those who grew up in states with lower legal drinking ages, not only drank less when they were under age 21, but also when they were age 21 to 25 (O'Malley & Wagenaar, 1991). However, it is not known whether efforts to reduce underage drinking lowers driving after drinking and alcohol related motor vehicle crash involvement later in life.

This national study of persons 18 and older, mean age 44, found that the earlier the age respondents started drinking, the more likely they were to report driving after drinking too much ever in their life and in the past year. Also the earlier the age of drinking onset, the greater the likelihood the respondents reported being in a motor vehicle crash because of their drinking, even after we analytically controlled for characteristics and behaviors associated with earlier drinking onset: whether or not respondents ever or currently had an alcohol dependence diagnosis, age, gender, education, race/ethnicity, marital status, smoking, or illicit drug use. A 4-fold increased alcohol-related crash risk for those who began drinking at age 14 relative to those who started after age 21 was found both ever in a respondent's life and during the year of the survey suggesting the relationship may not be purely a function of having more years of alcohol exposure. Also, the relationships were not solely a function of people who start drinking at an early age being more likely to become alcohol dependent. Among persons who were never alcohol dependent, relative to respondents who started drinking at age 21 or older, those who began drinking at age 14, and in each succeeding age up through age 20 were significantly more likely to report being in a motor vehicle crash after drinking ever and during the year prior to the survey.

Several methodological factors should be considered when interpreting the study's results. First, the distribution of responses on key drinking variables was similar in the NLAES sample when compared to other large national surveys. The National Longitudinal Survey of Labor Market Experience in Youth (NLSY)(N=12,686) was a follow up study of a survey designed to represent the non institutionalized population of young people age 14-21 as well as person 17-21 serving in the military in 1979. Grant, Harford and Stinson (In Review) reported the ages respondents in that sample started drinking. Compared to respondents in the NLAES who were in those age groups in 1979 (N=4210) the proportion of drinkers in the NLSY who started drinking at <14, and each consecutive year up to age 21 and older were virtually identical (data available upon request). When compared to the 1998 National Health Interview study, an area probability sample of over 100,000 persons nationwide, similar proportions of the NLAES sample were classified as ever having drunk alcohol 65% vs. 71% or as having met DSM IV alcohol dependence criteria 4.4% vs. 5.9%. In NLAES to be classified as ever drinking a person had to drink at least 12 drinks in a given year. In the National Alcohol Survey conducted in 1990 the proportion who met DSM IV dependence criteria was 3.9% (for data from the National Health Interview Study and the National Alcohol Survey see U.S. Department of Health and Human Services, 1997). The proportion who drove a motor vehicle after having too much to drink in the past year 5% in NLAES was similar to the percentage reporting they drove while impaired by alcohol in the past month 2.5% in CDC's 1993 national Behavioral Risk Factor Survey of over 102,263 adults 18 or older (Lui et.al., 1997).

Second, even though respondents said the motor vehicle crashes they experienced were because of their drinking, it is impossible to be certain that alcohol was the sole cause of the crash. However, alcohol use was present and the crash can be considered alcohol-related.

Third questions were not asked about the amount that respondents drove in general or crash involvement when they had not been drinking. Driving amounts could have influenced the key findings of this study. Also it would be of value to assess whether age of drinking onset was more closely related to alcohol related crash involvement than involvement in crashes where alcohol was not a factor.

Fourth, questions regarding driving after drinking and motor vehicle crash involvement after drinking could have been more objectively worded. In the NLAES survey, respondents were asked if they drove after drinking too much ever and in the past year and if they were in a motor vehicle crash because of their drinking ever and in the past year. A more objective approach would have been to ask respondents if they drove or were in a motor vehicle crash within an hour after consuming a specific number of drinks. The way questions were worded in the NLAES suggests the relations observed may have been stronger than reported. Persons who drive after drinking tend to believe they can consume more drinks and still drive safely than people who do not drive after drinking (National Highway Traffic Safety Administration, 1996). Consequently, they may under report the actual frequency with which they drove after drinking or were in a crash because of their drinking.

Fifth, there is the possibility that social desirability biases may have influenced respondents. Those least likely to admit drinking at an early age may be least likely to admit driving after drinking too much and being in a motor vehicle crash because of drinking.

Sixth, the survey was cross sectional and older adults may have difficulty remembering the age they first started to drink. Finally, because the survey was confidential, efforts were not made to link self-reports of crash involvement to driver records. Longitudinal studies asking adolescents when they first began to drink with more objective questions about driving after drinking and crash involvement and which explore actual traffic records need to be conducted.

The results of this study point to a need to explore why people who start drinking at an early age are more likely to report driving after drinking in the past year even among persons who are not alcohol dependent. Several factors may contribute to the relation. First, it is possible that people who engage in a variety of deviant or illegal behaviors at an early age are more likely to engage in several such behaviors later in life. However, the relationships in this study between early drinking onset and driving after drinking and motor vehicle crash involvement were independent of current or past use of illegal drugs or tobacco. Second, persons who start drinking at an early age more frequently drink heavily than those who start consuming alcohol later in life even if they are not alcohol dependent (Hingson, Heeren, Jamanka, & Howland, In Press). This could increase the likelihood that they would also drive more frequently after drinking. Also, persons who begin drinking at an early age may be less likely to believe that driving after drinking increases crash risks. They may believe for example, that the risk of traffic crashes and other injuries increase only for people who are visibly intoxicated. They may derive pleasure or a sense of self-esteem by taking driving risks. It is well known that people who drive after drinking are more likely to speed and are less likely to wear seat belts. Also, heavy consumption of alcohol may further impair the judgment of those who start drinking at an earlier age. After drinking they may be less likely to appreciate their increased crash risk than when sober. Whether those who start drinking at a younger age are as aware of traffic laws and as likely to believe those laws will be enforced also warrants exploration.

A recent report indicated that after a decade of decline, the percentage of high school seniors who drink and drive has increased the last two years (O'Malley & Johnston, 1999). The study reported here identifies another important reason to step up enforcement of the legal drinking age of 21 and to expand educational and community programs that delay the onset of drinking. The potential traffic safety benefits of delaying underage drinking may extend well beyond age 21.