This assumption is made explicit in a recent cost-effectiveness study of school-based drug prevention programs for reducing cocaine consumption (Caulkins et al., 1999). The authors found no studies that provided an estimate of the direct effect of prevention programs on later cocaine use. They had to estimate this effect indirectly in a two-stage process (from the National Household Survey of Drug Abuse), by combining estimates from evaluations of prevention program effects on the age of marijuana initiation and the correlation between the age of marijuana initiation and the quantity of cocaine later consumed. By combining these two estimates and making different assumptions about the permanence of the program effect on marijuana use, they arrived at a range of estimates of prevention’s effect on later cocaine use. According to these estimates, the percentage reduction in lifetime cocaine consumption due to prevention for a given cohort ranges from 2.9 to 13.6 percent, with a middle-range guess of 7.6 percent.
Of course, the validity of these estimates hinges on the assumption that the correlation between age at first use of marijuana and later cocaine use is due to certain individual propensities to use, and that prevention’s effect on marijuana use is due to its effect on this general propensity. Most important for the purpose of this report, the authors note a high degree of uncertainty that surrounds their estimates of the effects of prevention on later cocaine use—hence the title of their report: An Ounce of Prevention, A Pound of Uncertainty. More precise estimates of the effects of prevention on illegal drugs requires longitudinal follow-up of program participants and control groups in a large-enough sample to be able to detect mean group differences in very rare behaviors.
Aside from this paucity of data on the effects of prevention efforts on later illegal drug use, there are also differences across studies in the way tobacco, alcohol, and marijuana use are measured. Most studies use self-reports of drug use. Three types of self-report measures are generally used: prevalence, variety, and frequency measures. Prevalence measures assess status as a user in one’s life or during a certain time frame. Variety measures are counts of the number of different substances used and are often used to assess multiple drug use, which may be more dangerous than the use of a single substance. Frequency measures assess how often or how much an individual uses drugs.
These measures can be used to differentiate levels of use. For example, some studies have targeted drinking five or more drinks at one sitting as an outcome of interest. Other studies have developed cut-points to differentiate casual from heavy use, or varying degrees of drug involvement (e.g., Kellam et al., 1982). Age at first use is also sometimes measured, so that delays in onset of drug use can be assessed (e.g., Kellam and Anthony, 1998). Prevention studies generally have not assessed ef-