Cover Image


View/Hide Left Panel

drinking surveys and by concerns about the differing prevalence rates that result from the three major national surveys, with the NHSDA consistently reporting the lowest prevalence. The most recent and prominent effort, by Foster et al. (2003), used the NHSDA data to estimate the number of drinks consumed during the previous 30 days by current drinkers, based on multiplying together answers to questions on the number of drinking days during that period, and on how many drinks the respondent “usually” had on a drinking day.3 But NHSDA data were not used to estimate the participation rates—the proportions of the youth and adult populations that were current drinkers. Instead, the authors chose to use the Youth Risk Behavior Survey (YRBS) of the Centers for Disease Control and Prevention to estimate the proportion of underage individuals (12 to 20) who were current drinkers, and the Behavioral Risk Factor Surveillance Survey to estimate the proportion of adults who were current drinkers. These estimated participation rates were then combined with the NHSDA data on average drinks per current drinker to estimate the total amounts of alcohol consumed by youths and adults. The authors (Foster et al., 2003) calculated that underage individuals consumed 19.7 percent of the total number of drinks consumed in the United States in 1999, amounting to more than 830.6 million drinks per month. It is easy to find fault with the procedure adopted by Foster et al. (2003), but it is nonetheless not clear a priori whether the estimate is high, low, or about right. Three problems may be especially important. First, the average number of drinks consumed on drinking occasions is not well captured by the NHSDA item, which asks about the usual number of drinks. For example, a respondent who drinks a beer with supper every night and an additional six-pack on Saturday nights will “usually” drink one drink per occasion, but will drink an average of two drinks per occasion (14 for the week, divided by 7 days). What is not known is how each respondent interpreted this question and whether he or she then answered one or two drinks or something else. Whether this problem is greater for underage or adult drinkers is not clear, so the possible bias in the Foster et al. estimate could be either positive or negative.

Second, the YBRS does not seem well suited for estimating the participation rate for all youth aged 12 to 20 because the YRBS sampling frame is limited to youth aged 12 to 18 who are in school. Unlike the NHSDA, dropouts and older youths are not included in the YRBS. If the omitted groups have a higher drinking participation rate, as seems reasonable, then


Foster et al. (2003) excluded cases that reported 50 or more drinks as the usual number of drinks consumed on days they drank in the last 30 days because the response suggest a misunderstanding of the question. It is not clear to what extent inclusion of these responses in the analysis would have changed the outcomes.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement