Author Biographies: Biographical information of the authors is not made available in the research bulletin. A search online indicates that Carl McCurley shares a passion for analyzing and improving outcomes for court-involved children, youth and families that he was able to apply in the Models for Change program at the National Center for Juvenile Justice. McCurley, Director Administrative Office of the Courts/Washington State Center for Court Research (WSCCR). McCurley joined the Administrative Office of the Courts in 2007As the newly hired WSCCR director, he sought to broaden the Center's work, expanding beyond the customary focus of court operations to analyze the courts' impact on the lives of those they served.
Howard N. Snyder is considered to be the foremost expert on juvenile justice data, policy, and is now employed by the Bureau of Justice Statistics, U.S. Department of Justice in Washington D.C. (e-mail: Howard.- -- ).
Funding Source: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Delinquency Prevention
Problem: The problem is that levels of substance-related behaviors increase as youth age, such that, from the age of 12 to the age of 17, the self-reported use of alcohol 30 days prior to the research climbed more than eightfold.
Hypothesis: The research question was whether youth who have a single substance-related behavior are more likely to have other substance-related behaviors. HO: Youth who have a single substance-related behavior are more likely to have other substance-related behaviors. HA: Youth who have a single substance-related behavior are no more likely to have other substance-related behaviors.
Methodology: The research methodology was a retrospective meta-analysis of extant data from the National Longitudinal Study of Youth (NLSY97), in which a nationally representative sample of youth aged 12 to 17 years were interviewed regarding substance-related behaviors. This study modified the sample used in the two waves (1997 and 1998) of the National Longitudinal Study of Youth to produce nationally representative estimates of overall youth behavior and estimates of youth behavior within groups constructed according to age, gender, and race or ethnicity. These modifications were made by creating oversamples and then controlling the oversampling through weighted analysis, both commonly accepted statistical methodologies for addressing the natural variation in samples. To facilitate the analysis, the sample also underwent weighting that would establish equal proportions of youth for each age-year group. Included in this weighting was the adjustment to exclude the 17-year-olds who were disproportionately on the young side -- that is, showing an average age of 17 years and 3 months.
Analysis: The original data in the National Longitudinal Study of Youth was collected via interviews of more than 15,300 youth aged 12 to 17 years of age. The retrospective statistical method employed in this analysis is referred to as a meta-analysis. A meta-analysis is used to generate a more powerful estimate of the true effect size of a phenomenon than the estimate that is derived from a single inquiry under a given single set of conditions and assumptions.
The task force of the Organization for the Economic Cooperation and Development (OECD) charged with harmonizing the articulation of survey operation and technical design, particularly with regard to efficient sample design and weighting methodologies asserts that the methods employed in this meta-analysis of the National Longitudinal Study of Youth is, in fact, best practices. A 2005 report by this task force states:
"According to the report, the ideal survey would utilize probability methods for sample selection, adjust the results for differential non-responses, and weight the data so that the results are representative of the total reference population or volume of economic output for business sector surveys. The size of the sample should be determined by the width of the largest confidence internals allowable for key results. All surveys should clearly document the sampling frame, sample selection procedures, response rates, imputation methods for missing data and weighting." (Malgarini, 2005, p. 4).
The goals and objectives of the research were to provide information to individuals who are engaged in the identification and evaluation of youth who exhibit or report the type of substance-related behaviors that are the independent variables in the study. In other words, the research is intended to establish a base of knowledge for the professionals who make decisions about placement in programs, interventions and treatments, and the provision or delivery of services. The philosophical underpinning of the research is that substance-related behaviors, such as using marijuana, drinking alcohol, and buying or selling drugs, have substantive detrimental impact on and consequences for youth that it is essential to understand the co-occurrence of these substance-related behaviors in order to provide meaningful and effective interventions and support.
The authors established that validity is a moderate consideration for their study given that the findings do not align well with other substance abuse studies. The 30-day prevalence statistics -- how many youth exhibit substance-related behaviors in the population 30 days before responding to research questions -- fall somewhere between the findings derived from two major federally-funded surveys: The National Survey on Drug Use and Health (NLDUH) and Monitoring the Future (MTF). The prevalence findings for the National Survey on Drug Use and Health and the current study are similar. Prevalence for use of marijuana by youth age 12 to 17 years in the National Survey on Drug Use and Health (NLDUH) is 9% and for National Longitudinal Study of Youth (NLSY97), it is also 9%. Prevalence for use of alcohol by youth age 12 to 17 years in the National Survey on Drug Use and Health (NLDUH) is 21% and for National Longitudinal Study of Youth (NLSY97), it is 23%. The 30-day prevalence estimates in the Monitoring the Future (MTF) study were higher for both marijuana and alcohol than the finding in the National Survey on Drug Use and Health (NLDUH) and the National Longitudinal Study of Youth (NLSY97). The 30-day prevalence for use of marijuana in 10th graders in the Monitoring the Future (MTF) was 20% compared to 13% in the National Longitudinal Study of Youth (NLSY97). The 30-day prevalence for alcohol use by10th graders in the Monitoring the Future (MTF) was 40% compared to the estimate of 32% in the National Longitudinal Study of Youth (NLSY97).
No explanation is offered by the researchers regarding these differences in prevalence estimates for marijuana and alcohol use in the youth population, regardless of the fact that the differences are substantial. Sampling methods in the Monitoring the Future (MTF) consist of in-school administration of surveys, which tends to result in higher agreement to participate in research than other methods. The participants in Monitoring the Future (MTF) include adolescents, university and college students, and adults up through age 50 years of age. Since 1975, the study has been conducted by the University of Michigan's Institute for Social Research. Different sampling frames and sampling procedures can impact findings and the meta-analysis that is the focus of this discussion did not include data from the Monitoring the Future (MTF). Therefore, no modifications were made to the sample that would influence comparability and correct for disproportionality.
No ethical issues were created as a result of the meta-analysis. The participants in the National Longitudinal Study of Youth (NLSY97) agree to be interviewed at regular intervals and understand that their data is not individually identifiable and that their data is only presented in aggregate form. Consequently, no ethical issues with regard to confidentiality and participation surface in this meta-analysis. All such human subjects issues were addressed as part of the original research
Charts and graphs used: The visual and graphic display of data is highly satisfactory in this analysis. The visual displays of data are easy to interpret and effectively show relationships among the data. The data tables are complete, well-presented, and easy to understand. Each data table is comprehensively footnoted to refer to differences in subsamples based on age, gender, and race. Additionally, the authors have clearly indicated levels of confidence and "p" values to indicate significance.
Findings: The findings are reasonable and are based on empirically reliable methodologies. The findings are not remarkable and correspond to conclusions from other similar studies. For example, other studies have shown the correlation between substance-related behaviors, problem adolescent behavior, and juvenile delinquency. One notable difference, however, is in regard to the findings for African-American youth. The research indicates that compared to Hispanic or white youth, there is significantly less overlap of drug selling, alcohol use, and marijuana among African-American youth.
Implications of research: As with most federally-funded studies, there is the usual obfuscating Hispanic ethnicity category, in which participants self-identify themselves as being Hispanic or not being Hispanic, ultimately leading to nonsensical statements like this on federally-funded studies: "Racial categories do not include persons of Hispanic ethnicity. Youth of Hispanic ethnicity may not be of any race." This issue is not a direct reflection on the data collection, data analysis, and reporting of findings in this meta-analysis -- rather, it is commentary on the problem of…