which means that our measurement of trends should be affected very little by such biases” (Johnston et al., 1998:47–48). Similarly, Anglin et al. (1993) state that “it is easier to generate trend information…than to determine absolute level.”9

Consider, for instance, drawing inferences on the levels and trends in annual prevalence of use rates for adolescents during the 1990s.10 Data from MTF imply that annual prevalence rates for students in 12th grade increased from 29 percent in 1991 to 42 percent in 1997. Data from the NHSDA indicate that the annual prevalence rates of use for adolescents ages 12 to 17 increased from 13 percent in 1991 to 19 percent in 1997.11 Both series suggest that from 1991 to 1997, the fraction of teenagers using drugs increased by nearly 50 percent. Does the congruence in the NHSDA and MTF series for adolescents imply that both surveys identify the trends, if not the levels, or does it merely indicate that both surveys are affected by response problems in the same way?

This section discusses the effects of nonresponse and inaccurate response in the NHSDA and MTF on estimates of levels and trends in the prevalence of drug use. The conclusions vary by response problem. If nonresponse is the only significant problem (that is, incorrect responses do not occur), then the data provided by the NHSDA and MTF provide bounds on prevalence levels and identify the directions of sufficiently large changes in prevalence. If inaccurate responses are also present, then the data alone may not identify levels, trends, or even the directions of large changes in prevalence. With certain assumptions, both the levels and trends can be identified. In a paper prepared for the committee, Pepper provides a more detailed discussion (see Appendix D).

The committee concludes that these response problems, although not


These same ideas are expressed in the popular press as well. Joseph Califano, Jr., the former secretary of health, education and welfare, summarizes this widely accepted view about the existing prevalence measures: “These numbers understated drug use, alcohol and smoking, but statisticians will say that you get the same level of disassembling every year. As a trend, it’s probably valid” (Molotsky, 1999).


Annual prevalence measures indicate use of marijuana, cocaine, inhalants, hallucinogens, heroin, or nonmedical use of psychotherapeutics at least once during the year. Different conclusions about trends and levels might be drawn for other outcome indicators.


Similar qualitative differences in levels are generally found if one compares same-age individuals, although the magnitudes are much less extreme. Gfroerer et al. (1997b) reports that the age-adjusted prevalence rates from MTF are between 0.92 to 2.24 times the NHSDA rates. In all but one case (8th graders consuming cocaine), these ratios are over 1 with many reaching at least 1.4. The remaining differences may be due to variation in the survey methodologies as well as in the surveyed populations. MTF excludes dropouts, whereas the NHSDA surveys all adolescents living in noninstitutionalized quarters. In the NHSDA, adolescents may complete the questionnaire in the presence of their guardians. In MTF, guardians are not present, but peers are.

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