Objective. To compare early nondrinkers, experimenters, and drinkers on the prevalence of problem behaviors at grades 7 and 12 and at age 23 (N = 6338, 4265, and 3369, respectively).
Methods. Results are based on longitudinal self-report data from individuals who were originally recruited from 30 California and Oregon schools at grade 7 (1985) and assessed again at grade 12 (1990) and at age 23 (1995). Logistic regression was used to develop weighted estimates of the prevalence of academic difficulties, employment problems, substance use, and delinquent and violent behaviors within the 3 drinking status groups at grades 7, 12, and/or at age 23. Huber variance estimates, which adjust for weighting and clustering of observations, were used to assess the statistical significance of differences across groups.
Results. Early drinkers and experimenters were more likely than nondrinkers to report academic problems, substance use, and delinquent behavior in both middle school and high school. By young adulthood, early alcohol use was associated with employment problems, other substance abuse, and criminal and violent behavior.
Conclusions. Early drinkers do not necessarily mature out of a problematic lifestyle as young adults. Interventions for these high-risk youth should start early and address their other public health problems, particularly their tendency to smoke and use other illicit drugs.
- young adulthood
- academic problems
- employment problems
- substance use
National statistics point to the pervasiveness of underage drinking during both early and late adolescence. For example, it is estimated that half of adolescents have tried alcohol and one quarter have drunk to the point of intoxication by 8th grade.1 Adolescent alcohol use is one of the most important public health problems in the United States, playing a key role in the 4 leading causes of death among teens: motor vehicle accidents, unintentional injuries, homicides, and suicides.2 Further, adolescent drinking often co-occurs with other problem behaviors such as poor academic performance,3 absenteeism from school,4 violent and criminal behavior,5,6 illicit drug use,7 and risky sexual behavior.8,9 Each of these may impair development and pose significant challenges to making a successful transition from adolescence to adulthood.10
Although the evidence is strong that drinkers are more likely than nondrinkers to exhibit a wide range of other problem behaviors in adolescence, it is likely that there are subgroups of drinkers who differ in their risk for these problem behaviors. For example, 1 study identified 2 subgroups of drinkers, a problem-behavior prone group showing a rapid onset of alcohol use [reminiscent of the problem behavior syndrome described by Jessor and Jessor11], and a normally socialized group exhibiting a markedly lower tendency toward problem behaviors and relatively slow onset of drinking.12 Another found that moderate and problem drinkers differed in their behavioral regulation, with problem drinkers exhibiting more impulsivity, sensation seeking, and acting out.13 However, neither of these studies examined longer-term patterns of substance use or whether these groups continued to differ in other problem behaviors over time. A few longitudinal studies provide evidence that early drinking is associated with problem behaviors well into adulthood14–16; however, the scope, severity, and durability of these problems has not been well-documented.
In this study, we compare adolescent nondrinkers, experimenters, and drinkers on the prevalence of other problem behaviors over a 10-year period, at grades 7 and 12 and at age 23. We hypothesize that both experimenters and more committed drinkers will show a higher prevalence of problem behaviors compared with nondrinkers at grade 7, with drinkers being at highest risk. We expect a similar pattern in late adolescence and early adulthood. This research extends the literature by: including alcohol experimenters, as well as drinkers, in the analyses; comparing these groups on a wide range of behaviors, including substance use, academic and employment problems, violence, and delinquency; and doing so over the 10-year period from early adolescence to young adulthood. During this period, which covers the shift from middle school to high school and from high school to the early 20s, young people are particularly vulnerable to high-risk behavior.17,18
We used data from a longitudinal survey of 6527 individuals who were recruited from 30 California and Oregon schools at grade 7 (1985) and assessed again at grade 12 (1990) and age 23 (1995). The schools represented a wide range of community types, socioeconomic status, and racial/ethnic composition.19 Through extensive tracking and retention efforts, we retained 67% of the original baseline sample at grade 12 and 53% of the baseline sample at age 23. During grades 7 through 10, these tracking and retention efforts included scheduling make-up sessions for absentees, as well as collecting home address, new school, or third-party information on the 1000 to 2500 transferees, dropouts, and movers each year and contacting them through these alternate venues.20 At the grade 12 and young adult assessments, multiple methods were used to locate sample members and solicit their participation, including advance letters, Department of Motor Vehicle and national change of address checks, telephone tracking using locator forms and other sources, and participant payments. We excluded participants who were missing drinking information at grade 7 (n = 189). We also excluded individuals from the grade 12 and/or age 23 analyses if they dropped out of the study, were lost to follow-up, or otherwise failed to complete the relevant survey (final sample sizes: N = 6338 at grade 7; N = 4265 at grade 12; and N = 3369 at age 23). At baseline, 32% of the subsample was self-classified as a minority and 48% was female. High school dropouts (defined as someone who was not currently in school but did not have a high school diploma or general equivalency diploma certificate, or was enrolled in a continuation, alternative, or other special school), constituted 20% of our sample at the grade 12 assessment (weighted estimate).
Those who dropped out of the study or were lost to follow-up were more likely to have early behavioral problems compared with retainees21 and to be high school dropouts. To reduce attrition bias, we weighted the sample, using logistic regression to create predicted probabilities of responding to the 1990 and 1995 surveys that were derived from each respondent’s 7th grade information.22 These weights, which are the inverse of the predicted probabilities, remove 90% or more of the bias exhibited in the unweighted sample.23
At grade 7, we classified 23% of participants as nondrinkers (never had a drink, not even a few sips), 46% as experimenters (drank <3 times in the past year and not in the past month), and 31% as drinkers (drank 3 or more times in the past year or drank in the past month). Although we were unable to validate the self-reports of drinking, consistency checks across the first 4 waves of data collection indicated that the proportion of participants who denied using a particular substance after previously admitting use averaged about 5%.19 Further, physiological tests used to validate self-reported smoking showed that the vast majority (over 95%) told the truth about cigarette use.19 There is no reason to believe that self-reports of drinking should be more or less accurate. The lower percentage of nondrinkers in our sample at grade 7 compared with national samples of adolescents1 likely reflects our stricter definition of nondrinking (not even a few sips of alcohol).
School and Employment Problems
Indicators of school problems at grade 7 included skipping school or being sent out of the classroom more than once in the last year, missing 5 or more days of school during the current academic year, earning grades of C or worse, and ever repeating a grade. At grade 12, these indicators included skipping school, getting poor grades, repeating a grade, being suspended, and dropping out of school. At age 23, we focused on employment problems during the past year: skipping work for no good reason; receiving a serious warning on the job; and being fired.
Indicators of substance use at grade 7 included weekly smoking, weekly marijuana use, and any history of hard drug use. At grade 12, we added daily smoking, weekly alcohol use, binge drinking in the past month (defined as 5 or more drinks on a single occasion), alcohol abuse (pathological use combined with at least 1 indicator of social or occupational impairment), polydrug use in the past year, and experiencing multiple problems because of one’s alcohol and marijuana use.24 We defined hard drug use as any use during the past year. At age 23, we replaced the alcohol abuse item with a 3-item lifetime alcohol disorder screening instrument that focuses on binge drinking25 and 5 items from the Drug Abuse Screening Test.26 For each screener, participants were classified as abusers if they endorsed any items. We also broadened our assessment of multiple drug problems to include problems attributable to substances other than marijuana and added an item assessing whether or not participants had received any drug or alcohol treatment since age 18.
Other Problem Behaviors
At grade 7 we had 1 indicator of other problem behaviors—stealing from a store. At grade 12, indicators of other problem behaviors included past year stealing (from a store, a vehicle, something worth >$20, something worth $400 or more), selling marijuana or hashish, predatory violence (gang fighting, using force to get money or things from others, carrying a hidden weapon, attacking someone with the intent to seriously harm or kill), relational violence (hitting or threatening to hit family or nonfamily), felony commitment (buying/selling/holding stolen goods, taking a joy ride without the vehicle owner’s permission, breaking into property, arson or attempted arson), experiencing parenthood, and getting pregnant.24 At age 23 we dropped the parenthood and pregnancy items because we no longer considered them to be problem behaviors, as well as the relational violence items because they were in reference to household members only. Stealing from a store, taking a joy ride, and breaking into property were dropped, whereas ever arrested for drunk driving was added.
We used logistic regression to test for differences in the rate of problem behaviors across drinking status, computing Huber variance estimates to gauge the statistical significance of differences in means across groups. These means were unweighted for the grade 7 analysis, (which includes almost all of the 7th grade participants in the 30 schools), and weighted for the grade 12 and age 23 analyses. Huber variance estimates account for both weighting and the clustering of observations within schools and provide consistent standard errors under conditions of unequal variances in the clusters.27,28 Student t tests of logistic regression coefficients are reported for pairwise comparisons across groups. We used missing value imputation on all variables except grade 7 drinking status to deal with the small amount of missing data on most variables (ranging from <1% to 5% of cases) and to obtain consistent sample sizes across the analyses. This standard procedure involves using regression analysis to generate a predicted value (the criterion) from other variables that are strongly correlated with the criterion value, then substituting the predicted value for the missing value.
Additional analyses were conducted to assess the association between early drinking status and degree of involvement in problem behaviors. Composite measures of grade 12 and age 23 problem behaviors were created by summing the number of problems experienced by each participant from the list of indicator variables in Tables 2 and 3. Three variables were excluded from these composite measures because they were relevant to only 1 gender (pregnancy) or significantly overlapped with other variables (daily smoking and arrested for drunk driving). Scores on the composite measures ranged from 0 to 19 problems at grade 12 (mean [M]: 4.86; standard deviation: 3.98) and 0 to 18 problems at age 23 (M: 3.74; standard deviation: 3.50). A 1-way analysis of variance was conducted to compare the grade 7 drinking status groups on the number of problem behaviors in which they engaged at grade 12 and age 23. We also predicted the number of problem behaviors exhibited at grade 12 and age 23 from grade 7 drinking status alone, as well as from grade 7 drinking status plus demographic variables and each of the grade 7 problem behaviors shown in Table 1.
Grade 7 Profiles
Early drinkers were more likely than nondrinkers to report using other substances, stealing, and having school problems (except repeating a grade), with substance use showing the largest group differences (see Table 1). Compared with nondrinkers, early drinkers at grade 7 were: 19 times more likely to engage in weekly smoking and any hard drug use and 14 times more likely to engage in weekly marijuana use; 4.5 times more likely to steal; 3 times more likely to be sent out of class or skip class; 2 times more likely to be frequently absent from school; and 1.5 times more likely to have poor grades.
Early experimenters were also more likely to engage in problem behaviors compared with nondrinkers, but less likely to engage in such behaviors compared with early drinkers. Early experimenters at grade 7 were: 4 times more likely to engage in weekly smoking and 2.5 times more likely to have engaged in any hard drug use; 2 times more likely to steal compared with nondrinkers, and 1.2 to 1.7 times more likely to have school problems (except repeating a grade). Compared with experimenters, drinkers at grade 7 were: 20.5 times more likely to engage in weekly marijuana use, 8 times more likely to engage in any hard drug use, and 5 times more likely to engage in weekly smoking; 2 times more likely to steal; and 1.5 to 2 times more likely to be frequently absent, have poor grades, and exhibit behavioral problems in the classroom.
Grade 12 Profiles
As older adolescents, early drinkers continue to be more likely than nondrinkers to engage in substance use, have a variety of school problems, and exhibit delinquent and other problem behaviors. Although the differences are less pronounced at grade 12, some of the largest group differences remain in the area of substance use. Compared with nondrinkers, adolescent drinkers by grade 7 had the following characteristics at grade 12 (see Table 2). They were:
4 times more likely to be daily smokers,
3 to 5 times more likely to be illicit drug users and have multiple drug problems,
2.5 times more likely to be weekly or binge drinkers, be alcohol abusers, and have multiple alcohol problems,
1.3 to 2 times more likely to have school problems,
2 to 3.5 times more likely to steal, commit a felony, and sell drugs,
1.3 to 1.8 times more likely to engage in violence, and
1.5 to 2 times more likely to experience early parenthood and pregnancy.
Early experimenters also remain more likely than nondrinkers to exhibit problem behaviors as older adolescents. They were 1.5 to 2 times more likely to engage in substance use and have multiple substance use problems, and 1.2 to 1.7 times more likely to engage in delinquent behavior (except early parenthood or pregnancy) compared with nondrinkers. Early experimenters were also about 1.5 times more likely to have skipped class, earned poor grades, and been suspended. Drinkers continued to be at higher risk than experimenters on all variables except skipping class at grade 12.
Age 23 Profiles
As young adults, early drinkers were more likely than nondrinkers to have missed work for no good reason (but they were not more likely to have received a recent warning from a boss or been fired). They were also more likely to engage in substance use and to exhibit criminal and violent behavior (see Table 3). Compared with nondrinkers, young adults who were classified as drinkers by grade 7 were:
4 to 5 times more likely to be weekly marijuana users and show signs of drug abuse,
2 to 3 times more likely to engage in hard and polydrug use, have multiple drug problems, and have received any drug or alcohol treatment since age 18,
2.7 times more likely to be daily smokers,
1.7 to 2.3 times more likely to be weekly or binge drinkers, show signs of alcohol dependence, and have multiple alcohol problems,
about 1.5 times more likely to report missing work for no good reason,
4.5 times more likely to sell drugs,
2 to 3 times more likely to be arrested, and
2 times more likely to steal, commit a felony, or engage in predatory violence.
By young adulthood, early experimenters were more likely than nondrinkers to exhibit problem behaviors in certain areas, but not others. There were no significant differences between early experimenters and nondrinkers on employment-related problems, predatory violence, and being arrested for drunk driving. However, compared with grade 7 nondrinkers, the experimenters were about 1.5 to 2 times more likely to engage in various sorts of substance use and have multiple substance use problems, nearly 3 times more likely to engage in weekly marijuana use, and 1.5 times more likely to have received any drug or alcohol treatment since age 18. In addition, experimenters were 1.5 to 3 times more likely to engage in various sorts of criminal and violent behavior. By age 23, the early drinkers were at significantly higher risk compared with experimenters for all problems except receiving a job warning and stealing.
Predicting Subsequent Problem Behaviors From Grade 7 Drinking Status
The 3 drinking status groups significantly differed on the extent of problem behavior involvement (number of problem behaviors) at grade 12 (F(2,28) = 169.23; P < .001) and age 23 (F(2,28) = 130.24; P < .001), with posthoc comparisons indicating that all 3 drinking status groups significantly differed (all P values < .001). As expected, early nondrinkers had the fewest problem behaviors (M = 3.18 at grade 12; M = 2.31 at age 23), followed by early experimenters (M = 4.61 at grade 12; M = 3.56 at age 23) and early drinkers (M = 6.49 at grade 12; M = 5.09 at age 23). We conducted linear regression analyses to predict the number of grade 12 and age 23 problem behaviors from grade 7 drinking status, demographic characteristics, and other problem behaviors (see Table 4). Grade 7 drinking status was significantly associated with the number of problem behaviors at both follow-ups, explaining 10% of the variance at grade 12 and 9% of the variance at age 23. These associations remained significant after controlling for demographic characteristics and problem behaviors at grade 7.
Early adolescence has been described as a pivotal stage during which developmental trajectories diverge toward either healthy adjustment or psychopathology.29 Alcohol use may have an important influence on the lifecourse of adolescents to the extent that it affects their ability to achieve the developmental milestones of this period, such as the acquisition of key academic skills, establishment of positive peer relations, and formation of a strong self-concept.15,30 Developing profiles of early drinkers, experimenters, and nondrinkers may provide some insight into the trajectories of these groups through adolescence and early adulthood.
At grade 7, adolescent drinkers were more likely than experimenters and nondrinkers to be having school problems, using illicit drugs, and stealing. Early drinkers did not necessarily mature out of these problem behaviors by early adulthood. Ten years later, early drinkers remained at higher risk for involvement in a wide range of problem behaviors, including unreliable work attendance, substance use problems, violent behavior, and illegal activities. Although limited longitudinal research has found that early drinkers are more likely than other adolescents to subsequently exhibit certain problem behaviors, most notably heavy drinking,31–33 results from this study indicate that the scope of these differences reaches well beyond alcohol use. Further, the magnitude of the differences we have documented is often quite large, and many of these differences show up early and persist over time. In addition, this study shows that the high-risk profile associated with early alcohol use is not limited to more committed drinkers. Adolescents who had experimented with alcohol just once or twice during the past year at grade 7 were also more likely than nondrinkers to experience most types of academic, substance use, and delinquency problems during middle school and high school, and both drug abuse and other illegal activities as young adults.
Conduct problems at school and work, substance use and abuse, delinquency, criminal behavior, and violence are complex behaviors that are likely influenced by an array of biological, psychosocial, and cultural factors that interact in complex ways. Attempting to identify particular risk factors for problem behaviors that will occur years later in young adulthood, although important in terms of informing early intervention efforts, can be particularly challenging. In this context, it is impressive that grade 7 drinking status alone could account for 9% to 10% of the variance in the composite measures of problem behaviors assessed 5 years later at grade 12 and 10 years later at age 23. Further, these significant associations between early drinking status and later problem behavior remained after controlling for gender, race/ethnicity, age, parental education, family structure, and other types of early adolescent substance use (ie, cigarette, marijuana, hard drugs) and problem behaviors. Thus, results from this study suggest that early adolescent drinking should be considered among the important risk factors for problem behaviors during the transition to young adulthood.
Limitations of this study include the fact that our estimates of problem behaviors are based on a single sample of California and Oregon students; hence the extent to which the prevalence estimates based on this cohort would generalize to adolescents in other parts of the country is unknown. However, our previous work has indicated that prevalence rates for violence, drug use, dropping out of school, and multiple problem behaviors in this cohort during late adolescence are within the range typically found in national studies.21,23 Another limitation is the study’s exclusive reliance on self-report for information on drinking and other problem behaviors. Although it was not feasible to externally validate all self-reports of problem behaviors in this sample, reports of drug use in this cohort were found to be highly accurate when externally validated and subjected to internal consistency checks.34 Finally, the comparison of drinking status groups at grade 7 was limited by the lack of information in the dataset on problem behaviors other than substance use, academic problems, school behavior problems, and stealing. The strengths of the study include its focus on an array of different problem behaviors, the comparison of experimenters with both nondrinkers and more committed drinkers, the relatively long period of follow-up, and the substantial proportion of school dropouts in the sample.
Results from this study suggest possible candidates for identifying pathways through which early drinking might contribute to poor outcomes later in life: low academic achievement, erratic work history, escalated substance use and dependence, establishment of a criminal record, and exposure to a deviance-prone environment. Gaining a more sophisticated understanding of the long-term sequelae of early alcohol use is an important direction for future longitudinal research. Our findings also emphasize the importance of attempting to prevent drinking initiation in early adolescence.15 Nonetheless, many teens will have begun drinking by middle school. Efforts to curb alcohol misuse among high-risk youth may be more effective if they also address other problems these adolescents are likely to be experiencing, particularly other forms of drug use known to increase the risk of escalated drinking.7 Because alcohol use is already co-occurring with other problem behaviors by grade 7, such programs should be implemented by early adolescence.
This research was supported by grant AA10364 from the National Institute on Alcohol Abuse and Alcoholism.
- Received May 10, 2002.
- Accepted September 16, 2002.
- Reprint requests to (P.L.E.) RAND, 1700 Main St, Box 2138, Santa Monica, CA 90407-2138. E-mail:
- ↵US Department of Health and Human Services. Monitoring the Future National Survey Results on Drug Use, 1975–1999. Rockville, MD: National Institute on Drug Abuse; 2000
- ↵Jessor R, Jessor SL. Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth. New York, NY: Academic Press; 1997
- ↵Elliott DS, Huizinga D, Menard S. Multiple Problem Youth: Delinquency, Substance Use, and Mental Health Problems. New York, NY: Springer-Verlag Inc; 1989
- ↵Ellickson PL, Bell RM. Drug prevention in junior high: a multi-site longitudinal test. Science.1990;247 :1299– 1305
- ↵Huber PJ. The behavior of maximum likelihood estimates under non-standard conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability.1967;1 :221– 233
- ↵StataCorp. Stata Statistical Software, Release 6.0. College Station, TX: Stata Corporation; 1999
- ↵Newcomb MD. Psychosocial predictors and consequences of drug use: a developmental perspective within a prospective study. J Addict Dis.1997;16 :51– 89
- ↵Reinisch E, Bell RM, Ellickson PL. How Accurate Are Adolescent Reports of Drug Use? Santa Monica, CA: RAND; 1991. Publ. No. N-3198-CHF
- Copyright © 2003 by the American Academy of Pediatrics