Influence of a Nationwide Social Marketing Campaign on Adolescent Drug Use

Abstract

In this study, we examined whether awareness (recall) of the National Youth Anti-Drug Media Campaign (NYADMC) benefited youth by attenuating their drug use. Data were obtained from the National Survey of Parents and Youth (NSPY), an evaluative survey tool designed to monitor campaign progress over 4 years. A growth modeling strategy was used to examine whether change in message recall or campaign brand awareness was related to declining patterns of drug use. Two distinct growth trajectories were modeled to account for growth among younger (12 to 14) versus older (15 to 18) youth. Growth trajectories indicated steady and positive increases in alcohol, cigarette, and marijuana use over time. During the early portion of adolescence, youth reported more “brand” awareness, remembered more of the video clips depicting campaign messages, recalled more media stories about youth and drugs and more antitobacco ads, and reported more radio listening and less television watching. When they were older, these same youth reported declines in these same awareness categories except for specifically recalling campaign ads and radio listening. Models positing simultaneous growth in drug use and campaign awareness indicated mixed findings for the campaign. Overall early levels of campaign awareness had a limited influence on rates of growth, and in a few cases higher levels were associated with quicker acquisition of drug use behaviors. When they were younger, these youth accelerated their drug use and reported increasing amounts of campaign awareness. When they were older, increasing awareness was associated with declines in binge drinking and cigarette smoking. No effects for marijuana were significant but trended in the direction of increased awareness associated with declining drug use. The findings are discussed in terms of how they depart from previous reports of campaign efficacy and the potential efficacy of social marketing campaigns to reach a large and impressionable youthful audience with strategically placed advertisements.

Notes

¹Historically speaking, much of the “technology” of social marketing is based on the work of Lazarsfeld and Merton (1949) and Wiebe (1951–1952) and later refined by Kotler and Zaltman (1971; Kotler & Roberto, 1989). Social marketing campaigns use “brand” merchandising to sell or market social behavior change in a similar fashion as product marketing. In brand marketing schemes, information about a product is transmitted with the goal of changing behavior and increasing product consumption. Social marketing represents a variant of brand or commodity marketing, with the goal of persuasively changing behavior through value or attitudinal change. Standard campaign applications of these ideas generally have relied on public service announcements (PSAs) and ancillary communication strategies (e.g., television, radio, print media, billboards) to inform the public with a goal of changing beliefs, attitudes, and eventually behaviors. If diffusion of information about a product through merchandising will increase buying behavior, then transmission of information about social values may very well encourage people to change their behaviors. Examples of mass media interventions include campaigns to reduce cigarette smoking among youth (Flynn et al., 1992; Murray, Prokhorou, & Harty, 1994; Popham et al., 1994; Siegel & Bierner, 2000), reduce sexual risk, HIV and AIDS (e.g., Farr, Witte, Jarato, & Menard, 2005; McCombie, Hornik, & Arnarfi, 2002), child abuse (e.g., Stannard & Young), heart disease (Belicha & McGrath, 1990), and nutrition (Chew, Palmer, & Kim, 1998), to name a few.

²Detailed information on the scope of the media campaign, funding sources, and evaluation can be obtained at www.whitehousedrugpolicy.gov, http://www.drugabuse.gov/DESPR/Westat/ and through the ONDCP site http://www.mediacampaign.org/publications/index.html

³We are deeply grateful to the scholarly review that helped shape the contents of this article. Of the many points raised, one compelling one was the differences between treating awareness as a dependent measure and mapping changes in this measure to concurrent changes in drug use as opposed to treating awareness as if it was a manipulation. The different conceptualizations really attribute to whether we think of the campaign as a randomized trial in which there is a manipulation given equally to every participant. It is conceivable, for instance, that we think of the amount of recall each individual reports as the “manipulation” (reflecting their overall exposure), and this was assessed in the NSPY using industry standards as Gross Rating Points (GRPs). At an aggregate (zip code) level, we could measure the effect of GRPs on both recall and drug use, although there was little variation in the amount of campaign shown across the United States. The way we structured our analyses reinforces that the ultimate goal of the media campaign is to test whether the infusion of PSAs and antidrug messages shown through various media channels increases youth awareness that drugs will interfere with normal development and have negative consequences. Toward this end, the growth models express this emphasis by showing whether the trajectories of campaign awareness influence normative drug use trends. The models with awareness as a time-varying covariate express a slightly different and perhaps more experimental view.

⁴Variables modeled to obtain adjusted propensity scores included prior round measures that predict exposure (divided into quintiles) including school attendance, current grade level, academic performance (i.e., grades), participation in extracurricular activities, future plans, family functioning, antisocial behavior, association with antisocial peers, marijuana use by close friends, sensation seeking, self-reported tobacco or alcohol use of a long-standing nature, and amount of television viewing and radio listening. Inclusion of the television and radio measures in derivation of the CFP weight forced us to use nonresponse and selection weights in analyses of these exposure measures only.

⁵The adjustment or correction to variance estimators included replicate-specific poststratification factors. These were h_k = 2.57 for k = 1, 2,…, 60 replicate weights and h_k = 0.06 for k = 61, 62,…, 100 replicate weights. Calculation of the variance and standard deviation estimates is based on the formula: .

⁶The ACASI instrument was set up to include skip patterns for youth responding “No” to the drug experience questions (e.g., “Have you ever drank alcohol?”). Use of skip patterns resulted in many youth not responding to items tapping frequency or intensity of use. In order to include more youth in the analyses and examine variation attributed to nonuse, we added categories reflecting responses of “I never used…” to each question in the skip pattern. Adding these youth back into the analyses allowed us to estimate growth relations that consider the addition of youth transitioning from nonuse to use over time. (Otherwise, they would have been excluded from subsequent rounds and not analyzed.)

⁷Predictors in the attrition analyses included demographic and background measures (gender, dummy coded measures of race to contrast White, Black, and Hispanic versus all other ethnic groups, respectively, and a measure of religious service attendance); drug use measures (alcohol, cigarettes, and marijuana); campaign awareness (brand awareness and specific recall-aided exposure, television and radio exposure to antidrug messages, and television viewing and radio listening behavior), school-related factors (grades, absenteeism, educational plans, a summed index of extracurricular school activities, and whether the respondent attended school in the past 12 months); and measures of psychosocial risk (antisocial behavior and a regression-derived measure of risk for marijuana use).

¹The mean was freely estimated for Marijuana Use at age 18 for to improve the model fit.

²The mean was freely estimated for Specific Recall at age 18, and the variance for slope 1 was fixed at zero for proper model convergence.

³The variance for slope 1 was fixed at zero for Anti-Tobacco ads.

∗p < .05; ∗∗p < .01; ∗∗∗p < .001.

I = intercept; S₁ = slope 1; S₂ = slope 2; NA = not applicable due to fixed variance.

¹The variance was fixed at zero for the intercept (age 12) of the drug use variables.

²For Binge-Recall the mean at age 18 for Recall was freely estimated and the variance of slope 1 was fixed at zero.

³For Alcohol Use-Anti-Tobacco Ads, the variance of slope 1 for Anti-Tobacco Ads was fixed at zero.

⁴For Marijuana-Media Stories, the mean was freely estimated at age 18 for Marijuana.

∗p < .05; ∗∗p < .01; ∗∗∗p < .001.

DI = drug use intercept; DS = drug use slope; AI = awareness intercept; AS = awareness slope; NA = not applicable due to fixed variance.

⁸The column labeled AS₁→DS₂ that is contained in Table 3 reflects a “regression” of the slope for drug use (when the youth were older) on growth in campaign awareness when they were younger. This is another way to ask whether growth in awareness of campaign themes was protective and exerted an influence for younger youth as they matured. Only one relation is significant (r = .292, p ≤ .05) for alcohol and recalling seeing ads about youth and drugs. Nine of the 24 models tested did produce an inverse relation; however, these did not achieve statistical significance.

^a The variance was fixed at zero for the intercept (age 12) of the drug use variables.

A = Awareness; D = Drug Use; A12→D12 = Drug use for the cohort beginning at age 12 regressed on awareness (recall) at age 12.

∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

⁹Oregon Youth Substance Use Project (OYSUP), Funded by the National Institute on Drug Abuse (DA10767) with data collection running between 1999 and 2005, Principal Investigator Judy Andrews, Ph.D., Oregon Research Institute.

¹⁰Drug Abuse and Violence Prevention with Minority Youth, Funded by the National Institute on Drug Abuse (DA08905) with data collection running between 1998 and 2002, Principal Investigator, Gilbert J. Botvin, Ph.D., Weill Medical College of Cornell University, Institute for Prevention Research.