Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

As U.S. teen use of illicit street drugs, alcohol and cigarettes has declined since the mid 1990s, nonmedical prescription drug use has substantially increased. For instance, sedative use prevalence has risen among high school seniors from 2.8% in 1992 to 7.2% in 2005 (Friedman, New England Journal of Medicine, 6 April 2006, 1448-1450). Besides providing a recreational 'high', prescription drugs potentially aid sleep, enhance school performance and relieve stress, and are perceived as safer than street drugs. This trend is problematic, as many prescription drugs impede cognition and motor skills, are addictive, and might inflict long-term damage on the developing teenage brain.

This paper examines whether prescription and illicit drugs are substitutes for teens. Specifically, we investigate whether prescription drug use causally reduces illicit drug use. Teens might substitute prescription for illicit drugs as their availability from parents, friends, the internet and physicians expands. Instead, the observed negative relationship might simply stem from common unobserved factors. Heavier use of advertising by policymakers to emphasize the dangers of illicit drug use and pharmaceutical companies to sell their products could independently influence teen illicit and prescription drug consumption, as could corresponding evolution in societal attitudes. Similarly, the sets of teens who have become more likely to use prescription drugs and less likely to use illicit drugs might be distinct. In contrast, despite their opposing prevalence trends, prescription and illicit drugs might provide complementary pharmacological effects, so that the use of one increases, not decreases, that of the other. Or, certain teens might be prone to using either both or neither type of drug rather than one or the other. We analyze three data sets. The first two are the National Survey on Drug Use and Health (NSDUH, previously NHSDA) and the Monitoring the Future (MTF) study. Both provide annual data from 1990 onward, the former on individuals age 12 and above and the latter on 8th, 10th and 12th graders. The third is the College Alcohol Study (CAS), which surveyed college students in 1993, 1997, 1999 and 2001. These data allow for comparing the use of illicit and prescription drugs, both defined broadly and within specific categories (i.e. stimulants v. cocaine, tranquilizers v. heroin).

To identify causality, we estimate IV models using GMM. Our instruments are proxies for 'local' or 'peer' prescription drug use formed by aggregating reported prevalence by school/year (CAS), strata/year/cohort (NSDUH), or region/urbanicity/year/cohort (MTF). ?Strata? are groups of counties within states and ?cohort? represents teens or adults in the NSDUH and grade level in MTF, while prevalence is specific to drug category. We include analogously aggregated illicit drug use in the outcome equation to alleviate instrument endogeneity, for which we can test because the model is overidentified. Standard errors are clustered across groups used for instrument construction to correctly ascertain instrument strength.

Clarifying causality has implications regarding possible unintended consequences of anti-drug policies. Advertising that diminishes illicit drug consumption, for example, might require attention to prescription drug use to limit substitution from illicit to prescription drugs.