Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Feb 1.
Published in final edited form as: Psychol Addict Behav. 2018 Nov 8;33(1):26–34. doi: 10.1037/adb0000422

Alcohol, Tobacco, and Marijuana Expectancies as Predictors of Substance Use Initiation in Adolescence: A Longitudinal Examination

Kevin S Montes 1, Katie Witkiewitz 1, Matthew R Pearson 1
PMCID: PMC6367043  NIHMSID: NIHMS990014  PMID: 30407027

Abstract

Outcome expectancies have been found to be predictive of substance use. While development of expectancies may be dynamic during adolescence, it is unknown whether the rate of change (slope) in substance use expectancies is a risk factor for use onset across multiple substance use domains. The present study tested the hypothesis that the slope of positive and negative alcohol, tobacco, and marijuana use expectancies during mid-adolescence (9th-10th grade) would predict use onset of each respective substance during late adolescence (11th-12th grade). Data from 3,396 ethnically diverse high school students were collected across eight waves of assessment and analyzed within a latent growth modeling framework. Results revealed that the slopes of positive substance use expectancies amongst never-users of each respective substance predicted increased odds of onset (Alcohol: ORB = 7.73, p<.001; Tobacco: ORB = 5.58, p<.001; Marijuana: ORB = 2.49, p=.001). Only the slope of negative marijuana expectancies predicted increased odds of onset (Marijuana: ORB = .44, p=.04). Baseline level of positive and negative substance use outcome expectancies were also generally found to be associated with onset. For three common drugs used by adolescents, change in substance use expectancies during the first two years of high school may be a marker of risk propensity for substance use onset. Change in expectancies may be an important target in substance use prevention, with research indicating that expectancy challenge and life skills interventions being potentially efficacious.

Keywords: substance use expectancies, substance use onset, adolescence, latent growth modeling

Substance use onset can occur throughout an individual’s lifespan, however use onset that occurs during middle-to-late adolescence can have long-lasting negative health effects. For example, early use onset is associated with a greater frequency of future substance use, substance use-related problems, and an elevated risk of developing a substance use disorder (Chen, O’Brien, & Anthony, 2005; Ellickson, Martino, & Collins, 2004; Stueve & O’Donnell, 2005). Given the importance of early use onset in later risk for substance use problems and disorder, it is important to identify factors that could potentially prevent or delay substance use onset. The present study focuses on changes in substance use outcome expectancies as one potentially modifiable risk factor of substance use onset.

Outcome expectancies are beliefs regarding the positive and negative behavioral, cognitive, and emotional effects of a substance (Cooper, Kuntsche, Levitt, Barber, & Wolf, 2016). For example, positive substance use expectancies refer to the perceived positive effects of substance use, such as holding the belief that being under the influence of alcohol would make one feel enjoyment (Fromme, Stroot, & Kaplan, 1993). Negative substance use expectancies refer to the perceived negative effects of substance use, such as holding the belief that tobacco use would make one feel sick (Rohsenow et al., 2003). Substance use expectancies have consistently been found to be associated with alcohol (Jester et al., 2015; Montes et al., 2017), tobacco (Cohen, McCarthy, Brown, & Myers, 2002), and marijuana use (Flory, Lynam, Milich, Leukefeld, & Clayton, 2004). Specifically, cross-sectional and longitudinal research findings indicate that the level of positive substance use expectancies is positively associated with substance use behavior and use onset, whereas the level of negative substance use expectancies is negatively associated with substance use behavior and use onset (Doran et al., 2013). Importantly, substance use expectancy formation occurs during normative child and adolescent cognitive development, which often precedes substance use onset (Christiansen, Smith, Roehling, & Goldman, 1989; Miller, Smith, & Goldman, 1990). Change in substance use expectancies may be influenced by parental modeling of substance use (Brown, Tate, Vik, Haas, & Aarons, 1999), exposure to substance use information from the media (Boyle, LaBrie, Froidevaux, & Witkovic, 2016; Osberg, Billingsley, Eggert, & Insana, 2012), and personality factors (e.g., narcissism, sensation seeking; Gott & Hetzel-Riggin, 2018). Regardless of the antecedents of expectancy change, the perceived likelihood of positive outcomes occurring as a result of alcohol use tend to increase, whereas the perceived likelihood of negative outcomes tends to decrease, with age in early adolescence (Colder et al., 2014). Similar trends in tobacco and marijuana expectancies may also exist although less research has been conducted in these substance use domains (Trudeau, Spoth, Lillehoj, Redmond, & Wickrama, 2003).

Despite substance use expectancy formation being highly dynamic during youth (Colder et al., 2014; Dunn & Goldman, 1998), previous studies linking expectancies to use onset have predominately examined whether substance use initiation is predicted by prior static level of expectancies, overlooking whether expectancy change differentiates use onset risk. The rate of change in substance use expectancies may represent a risk factor for use onset above and beyond the static level of substance use expectancies. For example, in a seminal study, the rate of change (i.e., slope) of alcohol outcome expectancies was found to be fastest among participants who initiated alcohol use for the first time during the assessment period (Janssen, Treloar Padovano, Merrill, & Jackson, 2018). Moreover, decreases in negative alcohol outcome expectancies over time were found to be predictive of alcohol use onset indicating that within-person change in outcome expectancies may be associated with substance use onset. As an important extension of this prior work, the present study examined whether changes in positive and negative alcohol, tobacco, and marijuana expectancies during adolescence were prospective predictors of respective substance use onset. The present study is a secondary analysis of longitudinal data collected from a sample of high school students. Within a parallel process latent growth curve modeling framework, we hypothesized that the slope of positive and negative substance use expectancies during mid-adolescence (9th and 10th grade) would predict substance use onset during late adolescence (11th or 12th grade). The hypothesis that change in expectancies would be predictive of use onset was examined for three commonly-used drugs – alcohol, tobacco, and marijuana.

Method

Participants and Procedure

A total of 3,396 adolescents across ten data collection sites completed surveys that included questions assessing substance use expectancies and substance use. All students not enrolled in special education classes across the ten data collection sites were eligible to participate (N = 4,100). The majority of students assented to participate in the study (N = 3,874), of which, parent(s) provided consent (N = 3,396). In terms of attrition, 23 participants actively withdrew their enrollment in the study throughout the assessment period in addition to participants who were lost at follow up. For example, in terms of differential follow-up rates, completion rates ranged from 91% at wave 6 (completion rates at wave 7 and 8 were higher than at wave 6; wave 6 [N=3,078]) to 99% at wave 1 (N=3,383). The sample was 47% male and the age of participants at the first wave of assessment ranged from 13 to 15 years of age (M=14.08, SD=.40)1. The sample was predominantly Hispanic/Latino (47%), followed by Asian (16%), White (16%), Multiracial (6%), Other (6%), Black/African American (5%), Native Hawaiian/Pacific Islander (3%), and American Indian/Alaska Natives (1%). Data were collected as part of a longitudinal survey of substance use and mental health among high school students (Leventhal et al., 2015). The survey contained approximately 300 questions to assess approximately 250 outcomes across eight waves of assessment; thus, survey space was limited which precluded a more comprehensive assessment of substance use expectancies. There were eight waves of assessment, six months apart, starting in Fall 2013 when participants (9th grade) and ending in Spring 2017 (12th grade). Participants who completed surveys in person received small gifts (e.g., University of Southern California [USC] key chains) for completing each survey. Participants who moved and completed surveys over the phone or internet were given a $10 gift card for each survey completed. The USC institutional review board approved the study.

Measures

Substance Use Expectancies.

Positive and negative alcohol, tobacco, and marijuana expectancies were measured across each wave of the assessment period. Positive and negative expectancies were each measured with a single item for each substance (Positive: “I think I might enjoy, experience pleasure, or feel good using alcohol [tobacco/marijuana]”; Negative: “I think I might feel bad, sick, or embarrassed using alcohol [tobacco/marijuana]”) using a 4-point response scale (1=strongly disagree, 4=strongly agree). The items used to assess positive and negative substance use expectancies in the present study were adapted from the Cognitive Appraisal of Risky Events (Fromme, Katz, & Rivet, 1997) questionnaire. Moreover, the response format was adapted from a likelihood rating format to a degree of agreement rating format.

Substance Use Onset.

Substance use onset was measured across each wave of the assessment period. Participants were asked to indicate how old they were the first time they tried a substance and were given a list of substances that included “one full drink of alcohol [e.g., can of beer, glass of wine, wine cooler, or shot of liquor]”, “a few puffs of a cigarette [e.g., Marlboro, Camel, Newport, etc..]”, and “marijuana [pot, weed, grass, hash, reefer, or bud]”. The response options included a range of years, from “8 years or younger” to “17 years or older” as well as the response option, “I’ve never tried this substance”. In wave 1, participants who reported never having tried a substance received a recoded response of “0” with all other responses recoded as “1” to signify that a participant had initiated substance use. In subsequent assessments, participants were asked if they had ever used a specific substance in their life. Initiation (0=No, 1=Yes) during 11th or 12th grade was the primary outcome.

Data Analysis Plan

Frequency of substance use onset was analyzed using SPSS version 25 (IBM Corp, 2017). Mplus version 7.4 (Muthén & Muthén, 2014) was used for the growth curve analyses and to examine the correlations between binary outcome variables (i.e., tetrachoric correlations). To test the study’s hypotheses, data were linearly fitted within parallel process latent growth models to derive the intercepts and slopes from the first four waves of positive and negative expectancy data to examine whether these parameters were predictive of substance use onset that occurred during the last four waves of assessment (see Figure 1). Six models were estimated, one unadjusted (i.e., without covariates) and one adjusted (i.e., with covariates) model for each substance, and the latent intercept and slope parameters from the positive and negative expectancy trajectories were modeled simultaneously. Sex, age, ethnicity, living situation (i.e., number of parents residing at home), peer drinking, and parents’ level of college education were examined as covariates in the adjusted models as research indicates that they may be predictive of substance use onset (Chen & Jacobson, 2012; Cohen, Richardson, & LaBree, 1994; D’Amico & McCarthy, 2006; DuRant, Smith, Kreiter, & Krowchuk, 1999; Nelson, Van Ryzin, & Dishion, 2015; Nicolai, Moshagen, & Demmel, 2012; Patrick, Wightman, Schoeni, & Schulenberg, 2012). See Table 1 for coding of these covariates.

Figure 1.

Figure 1.

Open in a new tab

Model of the slope and intercept parameters from the latent growth model of positive and negative alcohol expectancies from the first half of the assessment period as predictors of alcohol onset in the second half of the assessment period (controlling for sex, age, ethnicity, family living situation, peer drinking, and parents’ education level).

Table 1.

Model Fit Statistics from Unadjusted and Adjusted Latent Growth Models

χ2(df) CFI (TLI) RMSEA (CI)
Unadjusted Model
Alcohol (Alc) 192.14(45)*** .92(.90) .044(.037-.050)
Cigarette (Cig) 198.82(45)*** .95(.93) .040(.034-.045)
Marijuana (Mj) 54.92(26)** .98(.97) .023(.014-.032)
Adjusted Model
Alcohol (Alc) 100.19(74)* .97(.97) .022(.001-.03)
Cigarette (Cig) 161.72(74)* .95(.94) .032(.026-.039)
Marijuana (Mj) 151.14(74)*** .92(.89) .032(.025-.040)
Open in a new tab

Notes. Indices of model fit were examined in both the unadjusted and adjusted models in the prediction of substance use initiation within each substance use domain. Unadjusted models consisted of models that did not control for the effects of covariates (e.g., sex [0=female, 1=male], age, ethnicity [0=Non-Hispanic White, 1=Hispanic], living situation [number of parents residing most of the time with the participants – coded 0=Mother and Father; 1=Other], peer drinking [how many of participants’ five closest friends use a substance], and parents’ level of education [0=parents did not complete some college; 1=one parent completed some college; 2=both parents completed some college]). Adjusted models consisted of models that controlled for the effects of covariates.

***

p<.001

**

p<.01

*

p<.05.

Only participants who did not initiate substance use onset during the first half of the assessment (i.e., first four waves) were included in the analyses (nalcohol=1773, ntobacco=2678, nmarijuana=2323). The analytic sample size for each model included participants with missing data related to substance use onset representing less than 8% of the sample in each model. Models were also conducted excluding participants with missing data related to substance use onset. Results from models excluding participants with missing data were not substantively different from the findings reported herein; thus, we present findings from the larger analytic subsamples. Attrition analyses were also conducted and indicated that missing alcohol, tobacco, and marijuana expectancy data were not associated with substance use initiation, ethnicity, or parents’ level of college education. Missing data were associated with participants’ sex and age, with being male (B =.35, p<.001) and older (B=.04, p=.03) associated with more missing alcohol expectancy data. For each model estimated, we used a pairwise missing data approach on the predictor variable(s) and the weighted least squares estimator (probit link function; Savolainen et al., 2018). The examination of model fit indices (e.g., χ2, CFI, TLI, RMSEA) based on recommended cutoffs (e.g., Kline, 2015) indicated good model fit for all unadjusted and adjusted models. For brevity, findings from the unadjusted models are presented in the text but all statistics from the unadjusted and adjusted models are presented in Table 1. Unstandardized effect estimates were presented (e.g., B and odds ratios [OR] with 95% confidence intervals). Statistical significance was evaluated at p<.05. In the examination of expectancies as a predictor of use onset, the intercept and slope of the expectancy trajectory each served as covariates in the model. Controlling for the other latent growth parameter in the prediction of use onset allowed for the examination of whether the expectancy slope explained unique variance in use onset not accounted for by baseline level of that expectancy. We also examined growth parameters as predictors of time-to-first use of a substance across waves. Results indicated there was not a significant effect of expectancy slopes in time-to-first-use of a substance.

Results

Descriptive Statistics

The timing of substance use onset across the assessment period is reported in Table 2. During the assessment period (9th-12th grade), differential rates of alcohol (50%), tobacco (30%), and marijuana (46%) use onset were observed. In Table 3, an examination of mean expectancy endorsement across the entire assessment period revealed a linear trend (positive slope for positive expectancies and negative slope for negative expectancies). In Table 4, a statistically significant negative correlation between the positive and negative expectancy slopes within each domain was evidenced. The outcome variables were also found to be positively correlated with each other (alcohol-tobacco initiation: r=.51, p<.001; alcohol-marijuana: r=.70, p<.001; tobacco-marijuana: r=.63, p<.001).

Table 2.

Timing of Substance Use Onset Across Assessment Waves

Alcohol Tobacco Marijuana Dual Onset*
At or before Fall 9th grade, N(%) 864(25%) 348(10%) 503(15%) 484(14%)
Spring 9th Grade, N(%) 327(10%) 162(5%) 226(7%) 127(4%)
Fall 10th Grade, N(%) 241(7%) 109(3%) 154(5%) 74(2%)
Spring 10th Grade, N(%) 180(5%) 88(3%) 179(5%) 58(2%)
Fall 11th Grade, N(%) 123(4%) 72(2%) 125(4%) 37(1%)
Spring 11th Grade, N(%) 506(15%) 350(10%) 563(17%) 405(12%)
Fall 12th Grade, N(%) 180(5%) 130(4%) 183(5%) 83(2%)
Spring 12th Grade, N(%) 123(4%) 100(3%) 150(4%) 57(2%)
No Initiation (or no dual initiation), N(%) 841(25%) 2026(60%) 1302(39%) 2060(61%)
Open in a new tab

Notes. %= Percentage of participants who initiated or did not initiate substance use during each wave of assessment (percentages may not add to 100% due to rounding).

*

Percentages for dual onset represent a participant having endorsed the initiation of at least two substances during an assessment wave.

Table 3.

Mean Latent Growth Parameters and Mean Positive and Negative Substance Use Expectancies Across Waves 1–8

Estimated Parameters Waves 1 – 8
Intercept Slope Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8
M(SE) M(SE) M(SD) M(SD) M(SD) M(SD) M(SD) M(SD) M(SD) M(SD)
Alcohol
PAE 1.89(.02)*** .07(.01)*** 1.79(.89) 1.92(.99) 1.97(1.01) 2.05(1.05) 2.15(1.06) 2.16(1.08) 2.20(1.08) 2.30(1.08)
NAE 2.78(.03)*** −.09(.01)*** 2.85(1.03) 2.72(1.10) 2.67(1.10) 2.52(1.12) 2.60(1.08) 2.53(1.09) 2.46(1.09) 2.44(1.06)
Tobacco
PTE 1.27(.02)*** .02(.01) 1.25(0.56) 1.28(0.62) 1.30(0.65) 1.32(0.67) 1.33(0.68) 1.32(0.67) 1.33(0.68) 1.36(0.71)
NTE 3.25(.03)*** −.07(.01)*** 3.26(1.01) 3.16(1.11) 3.15(1.11) 3.00(1.18) 3.11(1.11) 3.06(1.13) 3.01(1.14) 2.99(1.14)
Marijuana
PME 1.67(.04)*** .07(.02)*** 1.61(0.94) 1.76(1.04) 1.80(1.08) 1.83(1.10) 1.90(1.11) 1.93(1.12) 1.99(1.14) 2.07(1.15)
NME 3.05(.03)*** −.10(.01)*** 3.09(1.11) 2.90(1.17) 2.88(1.18) 2.74(1.23) 2.77(1.18) 2.69(1.19) 2.64(1.18) 2.57(1.17)
Open in a new tab

Notes. M=Means. SE=Standard Error. SD=Standard Deviation. PAE=Positive Alcohol Expectancies, NAE=Negative Alcohol Expectancies; PTE=Positive Cigarette Expectancies, NTE=Negative Cigarette Expectancies; PME=Positive Marijuana Expectancies, NME=Negative Marijuana Expectancies. Unstandardized estimates from unadjusted models are provided. The response scale for the positive (e.g., “I think I might enjoy, experience pleasure, or feel good using alcohol”) and negative (“I think I might feel bad, sick, or embarrassed using alcohol”) expectancy questions were: 1=Strongly Disagree, 2=Disagree, 3=Agree, 4=Strongly Agree (Disagree > 2.5 > Agree).

***

p<.001

**

p<.01

p<.05.

Table 4.

Correlations Between Growth Parameters in Parallel Process Latent Growth Model of Substance Use Expectancies

Latent Growth Parameters
Alcohol Expectancies (Wave 1–4) I1 (SE) S1 (SE) I2 (SE) S2 (SE)
I1 (PAE) -- −.18(.07)* −.79(.03)*** .37(.07)***
S1 (PAE) −.04(.02) -- .14(.06)* −.61(.08)***
I2 (NAE) −.41(.05)*** .01(.01) -- −.33(.08)***
S2 (NAE) .06(.02)*** −.03(.01)*** −.06(.02)* --
Cigarette Expectancies (Wave 1–4 ) I3 S3 I4 S4
I3 (PCE) -- −.22(.04)*** −.70(.04)*** .23(.07)**
S3 (PCE) −.03(.01)*** -- .12(.05)* −.52(.08)***
I4 (NCE) −.19(.02)*** .02(.01)* -- −.08(.11)
S4 (NCE) .02(.01)* −.01(.01)*** −.01(.02) --
Marijuana Expectancies (Wave 1–4) I5 S5 I6 S6
I5 (PME) -- −.26(.06)*** −.84(.03)*** .32(.07)***
S5 (PME) −.07(.03)* -- .22(.06)*** −.71(.08)***
I6 (NME) −.60(.06)*** .06(.02)*** -- −.29(.08)***
S6 (NME) .08(.02)*** −.06(.01)*** −.10(.03)*** --
Open in a new tab

Notes. Correlation coefficients above the diagonal were estimated from the unadjusted models whereas coefficients below the diagonal were estimated from the adjusted models. I1=Positive alcohol expectancies (PAE) intercept, S1=Positive alcohol expectancies slope; I2=Negative alcohol expectancies (NAE) intercept; S2=Negative alcohol expectancies slope. I3=Positive cigarette expectancies (PCE) intercept; S3=Positive cigarette expectancies slope; I4=Negative cigarette expectancies (NCE) intercept; S4=Negative cigarette expectancies slope. I5=Positive marijuana expectancies (PME) intercept; S5=Positive marijuana expectancies slope; I6=Negative marijuana expectancies (NME) intercept; S6=Negative marijuana expectancies slope. SE=Standard Error. Standardized estimates are provided.

***

p<.001

**

p<.01

*

p<.05.

Change in Substance Use Expectancies as a Predictor of Substance Use Onset

Positive expectancies.

The slopes of positive expectancies from the unadjusted models were found to be statistically significant prospective predictors of substance use onset across all three substances (see Table 5). Specifically, the slopes from the alcohol (β=2.04, p<.001; ORB = 7.73), tobacco (β=1.72, p<.001; ORB = 5.58), and marijuana (β=.91, p=.001; ORB = 2.49) expectancy trajectories during 9th and 10th grade (Waves 1–4) were positively associated with their respective substance use onset in 11th or 12th grade (Waves 5–8). The ORB term can be interpreted as an increase in the odds of initiating substance use for the first time per 1-unit increase in positive expectancies slope. The intercepts of positive substance use expectancies were also found to be statistically significant prospective predictors of use onset. Specifically, the intercepts of the positive alcohol (β=.64, p<.001; ORB = 1.90), tobacco (β=.78, p<.001; ORB = 2.18), and marijuana (β=.59, p<.001; ORB = 1.81) expectancy trajectories during 9th and 10th grade were positively associated with their respective substance use onset in 11th or 12th grade.

Table 5.

Latent Growth Parameters and Covariates as Predictors of Substance Use Onset

Alcohol
Unadjusted Adjusted
B(SE) OR (95% CI) B(SE) OR (95% CI)
PAE Intercept .64(.14)* 1.90(1.40–2.41)* .83(.12)* 2.29(1.75–2.82)*
PAE Slope 2.04(.43)* 7.73(1.2–14.3)* 1.61(.29)* 4.99(2.12–7.87)*
NAE Intercept −.08(.12) .92(.70–1.15) −.21(.14) .98(.72–1.24)
NAE Slope .46(.43) 1.59(−.15–3.32) −.31(.50) .73(.03–1.44)
Sex - - −.38(.11)* .68(.53-.83)*
Age - - .09(.14) 1.10(.79–1.40)
Ethnicity - - .05(.13) 1.06(.78–1.33)
Living Situation - - −.01(.12) .99(.77–1.21)
Peer Drinking - - −.03(.04) .97(.90–1.04)
Parents' Education - - −.01(.06) .98(.86–1.10)
Tobacco
Unadjusted Adjusted
B(SE) OR (95% CI) B(SE) OR (95% CI)
PTE Intercept .78(.17)* 2.18(1.45–2.91)* .84(.29)* 2.32(1.02–3.62)*
PTE Slope 1.72(.33)* 5.58(1.95–9.20)* 2.67(.69)* 14.84(−5.24–34.92)
NTE Intercept −.33(.13)* .72(.53-.91)* −.39(.25) 2.32(1.02–3.62)
NTE Slope −.39(.42) .68(.12–1.23) .26(.83) 1.30(−.81–3.42)
Sex - - .13(.08) 1.13(.95–1.32)
Age - - .28(.11)* 1.32(1.04–1.59)*
Ethnicity - - .01(.01) 1.00(.82–1.19)
Living Situation - - −.01(.10) 1.00(.83–1.67)
Peer Drinking - - .05(.04) 1.05(.97–1.34)
Parents' Education - - .01(.05) 1.01(.90–1.10)
Marijuana
Unadjusted Adjusted
B(SE) OR (95% CI) B(SE) OR (95% CI)
PME Intercept .59(.10)* 1.81(1.45–2.17)* .68(2.12) 1.96(−6.39–10.31)
PME Slope .91(.28)* 2.49(1.13–3.85)* 1.17(16.81) 3.22(−103.09–109.54)
NME Intercept −.13(.11) .88(.69–1.07) −.13(.33) .88(−4.87–6.63)
NME Slope −.81(.39)* .44(.11-.78)* −.61(6.44) .54(−6.33–7.42)
Sex - - .04(.08) 1.04(.87–1.21)
Age - - .30(.10)* 1.34(1.08–1.62)
Ethnicity - - .10(.10) 1.01(.89–1.31)
Living Situation - - −.30(55.09) 1.01(−.24–2.27)
Peer Drinking - - −.01(1.64) 23.14(9.14–37.15)
Parents' Education - - −.005(.05) .98(.88–1.07)
Open in a new tab

Notes. PAE = Positive Alcohol Expectancies, NAE = Negative Alcohol Expectancies, PTE = Positive Tobacco Expectancies, NTE = Negative Tobacco Expectancies, PME = Positive Marijuana Expectancies, NME = Negative Marijuana Expectancies. Unadjusted = model without covariates, Adjusted = model with covariates. B = Unstandardized beta coefficient, SE = Standard Error, OR = Unstandardized Odds Ratio, 95% CI = 95% Confidence Interval (Lower Bound-Upper Bound). In each model, all latent growth parameters (e.g., intercepts and slopes of positive and negative expectancies) were derived from wave 1–4 data and were evaluated in the same model in the prediction of substance use onset that occurred in wave 5–8 (0=no initiation during wave 5–8; 1 = initiated during wave 5–8). Covariates include: sex [0=female, 1=male], age, ethnicity [0=Non-Hispanic White, 1=Hispanic], living situation [number of parents residing most of the time with the participants – coded 0=Mother and Father, 1=Other], peer drinking [how many of participants’ five closest friends use a substance], and parents’ level of education [0=parents did not complete some college; 1=one parent completed some college; 2=both parents completed some college]).

*

p <.05

Negative expectancies.

Only the slope of negative marijuana expectancies from the unadjusted models was found to predict marijuana use onset (β=−.81, p=.04; ORB = .44: see Table 5). Only the intercept of negative tobacco expectancies was found to be statistically significant prospective predictor of use onset (β=−.33, p=.01; ORB = .72). The intercepts of negative alcohol and marijuana expectancies were not statistically significant predictors of respective use onset.

Discussion

The present study focused on the examination of the change in positive and negative substance use expectancies during 9th and 10th grade as a predictor of substance use onset during 11th or 12th grade. The purpose of the study was to extend prior research (e.g., Janssen et al., 2018) by elucidating whether change in positive and negative alcohol, tobacco, and marijuana expectancies were a risk factor for substance use onset in high school. Consistent with our hypothesis, the observed increase in positive substance use expectancies over time prospectively predicted a higher likelihood of substance use onset across all substance use domains (except in the adjusted marijuana model where poor model fit was exhibited). Given that research indicates that positive substance use expectancies may increase with age during childhood and adolescence (Colder et al., 2014; Miller et al., 1990; Wahl, Turner, Mermelstein, & Flay, 2005), and that results from the present study indicated that changes in positive substance use expectancies was a risk factor for substance use onset, research is needed to examine antecedents of change in substance use expectancies in adolescence. Importantly, a different constellation of antecedents of expectancy change may exist depending on the substance under observation, with exposure to substance use information from the media (Boyle et al., 2016) being potential antecedents of expectancy change to target in future research [not measured in the study]. In addition, dynamic changes in e-cigarette and marijuana use expectancies among adolescence need to be monitored and examined as predictors of use onset given the rise of e-cigarette use and public discourse surrounding marijuana use legalization (Harrell et al., 2014; Volkow, Baler, Compton, & Weiss, 2014). Such investigations would be guided by expectancy (Bandura, 1986) and social learning theory (Rotter, Chance, & Phares, 1972) where indirect experiences with substance use (e.g., parental modeling and media influence) are posited to inform the development of substance use expectancies.

Only the slope of negative marijuana expectancies was found to be associated with use onset, extending previous findings in the alcohol literature (e.g., Janssen et al., 2018). However, the slopes associated with negative substance use expectancies were largely found to not be significant prospective predictors. Although largely inconsistent with our hypothesis, research examining positive and negative alcohol expectancies have generally found more robust effects for positive alcohol expectancies as predictors of substance use among younger populations (Zamboanga, Horton, Leitkowski, & Wang, 2006). Thus, it may be that expectancy change and its effect on substance use onset may be more salient for positive (relative to negative) expectancies.

Implications for Intervention Efforts

The finding that change in positive substance use expectancies is a predictor of use onset across three substance use domains, over and above static level of expectancies, represents a novel and important extension to the expectancy and substance use literature. Given the findings from the present study, expectancy challenge interventions – which attempt to challenge positive substance use expectancies that individuals hold and that have been found to be effective at reducing substance use (Wood, Capone, Laforge, Erickson, & Brand, 2007) – may need to be adapted to incorporate an awareness component so that individuals learn to be more attuned to shifting expectancies. Rather than merely administering a preventive intervention addressing beliefs about drugs at the outset of high school, continued challenging of substance use expectancies throughout the first half of high school may be more beneficial to slow the acceleration of positive substance use expectancies formation and in turn help to delay substance use onset. In addition, a school-based life skills intervention administered to participants in middle school was found to be associated with a slower rate of decrease in negative substance use expectancies (Trudeau et al., 2003). Given that change in negative marijuana expectancies was predictive of marijuana use onset, a life skills intervention may be effective in delaying adolescent marijuana use onset. Moreover, attempts to tailor a life skills intervention to target positive substance use expectancies may also be effective at delaying substance use onset more broadly.

Limitations and Conclusions

Study limitations should be noted. First, for each substance, only two items were used to assess positive and negative substance use expectancies, and these items focused on relatively narrow construct domains (i.e., pleasure, enjoyment, positive affect, sick, embarrassment). Thus, construct coverage and reliability of the measures may be suboptimal. While the items do reflect the measurement of outcome expectancies as conceptualized in expectancy theory (Alfonso & Dunn, 2007; Copeland, Brandon, & Quinn, 1995; Ham, Stewart, Norton, & Hope, 2005), single item indicators may lead to biased estimation. Future research efforts are strongly encouraged to use psychometrically valid measures of substance use outcome expectancies with better construct converge that also contain subscales to provide more specific details related to the types of expectancies that should be targeted in interventions. Lastly, study hypotheses did not focus on the earliest initiators of substance use in order to separate distinct periods of pre-onset expectancies and use onset risk. Additional research is needed to determine if change in expectancy trajectories is similarly related to use onset among early initiators.

In sum, in the examination of three common drugs of abuse, change in positive substance use expectancies during the first two years of high school were found to be a marker of risk propensity for substance use onset, independent of risk explained by the static expectancy level at high school outset. Change in positive expectancies may be an important target in adolescent substance use prevention and risk assessment. Additional research is needed to translate findings from the present study to improve expectancy-based interventions for youth substance use.

Supplementary Material

Table

Acknowledgements:

The research reported in this manuscript was supported by grants from the National Institutes of Health (NIH; R01-DA033296, P50-CA180905). Manuscript preparation was also supported by grants from the NIH (T32-AA018108 [Montes; PI McCrady]; K01-AA023233 [Pearson]). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. The authors report that they have no financial conflicts of interest with respect to the content of this manuscript.

Footnotes

1

The original sample included participants who were 12 (n=1) and 16 (n=10) years of age. Given their low frequency in the sample (see Supplemental Table 1 for age frequencies in the full and age-restricted sample), we used an age-restricted sample so that study findings would be more generalizable to the typical age range of incoming high school freshman students. We also considered examining age-distributed rather than grade-distributed data. Given the low variability in participants’ baseline age in the age-restricted sample (~82% of participants were 14 years of age), and that the study design and assessment schedule was based on grade level, we wanted to use an analytic approach consistent with the study design. Thus, we elected to examine grade-distributed rather than age-distributed data.

Prior Dissemination: Prior dissemination of the data occurred at the Research Society on Alcoholism and Research Society on Marijuana conferences in the form of poster presentations.

References

  1. Alfonso J, & Dunn ME (2007). Differences in the marijuana expectancies of adolescents in relation to marijuana use. Substance Use & Misuse, 42(6), 1009–1025. [DOI] [PubMed] [Google Scholar]
  2. Bandura A (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall, Inc. [Google Scholar]
  3. Boyle SC, LaBrie JW, Froidevaux NM, & Witkovic YD (2016). Different digital paths to the keg? How exposure to peers' alcohol-related social media content influences drinking among male and female first-year college students. Addictive Behaviors, 57, 21–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Brown SA, Tate SR, Vik PW, Haas AL, & Aarons GA (1999). Modeling of alcohol use mediates the effect of family history of alcoholism on adolescent alcohol expectancies. Experimental and Clinical Psychopharmacology, 7(1), 20–26. [DOI] [PubMed] [Google Scholar]
  5. Chen C-Y, O’Brien MS, & Anthony JC (2005). Who becomes cannabis dependent soon after onset of use? Epidemiological evidence from the United States: 2000–2001. Drug and Alcohol Dependence, 79(1), 11–22. [DOI] [PubMed] [Google Scholar]
  6. Chen P, & Jacobson KC (2012). Developmental trajectories of substance use from early adolescence to young adulthood: Gender and racial/ethnic differences. Journal of Adolescent Health, 50(2), 154–163. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Christiansen BA, Smith GT, Roehling PV, & Goldman MS (1989). Using alcohol expectancies to predict adolescent drinking behavior after one year. Journal of Consulting and Clinical Psychology, 57(1), 93–99. [DOI] [PubMed] [Google Scholar]
  8. Cohen DA, Richardson J, & LaBree L (1994). Parenting behaviors and the onset of smoking and alcohol use: a longitudinal study. Pediatrics, 94(3), 368–375. [PubMed] [Google Scholar]
  9. Cohen LM, McCarthy DM, Brown SA, & Myers MG (2002). Negative affect combines with smoking outcome expectancies to predict smoking behavior over time. Psychology of Addictive Behaviors, 16(2), 91–97. [DOI] [PubMed] [Google Scholar]
  10. Colder CR, O'Connor RM, Read JP, Eiden RD, Lengua LJ, Hawk LW Jr, & Wieczorek WF (2014). Growth trajectories of alcohol information processing and associations with escalation of drinking in early adolescence. Psychology of Addictive Behaviors, 28(3), 659–670. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Cooper ML, Kuntsche E, Levitt A, Barber LL, & Wolf S (2016). Motivational models of substance use: A review of theory and research on motives for using alcohol, marijuana, and tobacco. The Oxford Handbook of Substance Use and Substance Use Disorders, 1, 1–117. [Google Scholar]
  12. Copeland AL, Brandon TH, & Quinn EP (1995). The Smoking Consequences Questionnaire-Adult: Measurement of smoking outcome expectancies of experienced smokers. Psychological Assessment, 7(4), 484–494. [Google Scholar]
  13. D’Amico EJ, & McCarthy DM (2006). Escalation and initiation of younger adolescents’ substance use: The impact of perceived peer use. Journal of Adolescent Health, 39(4), 481–487. [DOI] [PubMed] [Google Scholar]
  14. Doran N, Khoddam R, Sanders PE, Schweizer CA, Trim RS, & Myers MG (2013). A prospective study of the Acquired Preparedness Model: The effects of impulsivity and expectancies on smoking initiation in college students. Psychology of Addictive Behaviors, 27(3), 714–722. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Dunn ME, & Goldman MS (1998). Age and drinking-related differences in the memory organization of alcohol expectances in 3rd-, 6th-, 9th-, and 12th-grade children. Journal of Consulting and Clinical Psychology, 66(3), 579–585. [DOI] [PubMed] [Google Scholar]
  16. DuRant RH, Smith JA, Kreiter SR, & Krowchuk DP (1999). The relationship between early age of onset of initial substance use and engaging in multiple health risk behaviors among young adolescents. Archives of Pediatrics & Adolescent Medicine, 153(3), 286–291. [DOI] [PubMed] [Google Scholar]
  17. Ellickson PL, Martino SC, & Collins RL (2004). Marijuana use from adolescence to young adulthood: Multiple developmental trajectories and their associated outcomes. Health Psychology, 23(3), 299–307. [DOI] [PubMed] [Google Scholar]
  18. Flory K, Lynam D, Milich R, Leukefeld C, & Clayton R (2004). Early adolescent through young adult alcohol and marijuana use trajectories: Early predictors, young adult outcomes, and predictive utility. Development and Psychopathology, 16(1), 193–213. [DOI] [PubMed] [Google Scholar]
  19. Fromme K, Katz EC, & Rivet K (1997). Outcome expectancies and risk-taking behavior. Cognitive Therapy and Research, 21(4), 421–442. [Google Scholar]
  20. Fromme K, Stroot EA, & Kaplan D (1993). Comprehensive effects of alcohol: Development and psychometric assessment of a new expectancy questionnaire. Psychological Assessment, 5(1), 19–26. [Google Scholar]
  21. Gott AJ, & Hetzel-Riggin MD (2018). What Did You Expect? Substance Use Expectancies Mediate the Relationships Between Dark Triad Traits, Substance Use, and Substance Preference. Psychological Reports, 1–22. [DOI] [PubMed] [Google Scholar]
  22. Ham LS, Stewart SH, Norton PJ, & Hope DA (2005). Psychometric assessment of the Comprehensive Effects of Alcohol Questionnaire: Comparing a brief version to the original full scale. Journal of Psychopathology and Behavioral Assessment, 27(3), 141–158. [Google Scholar]
  23. Harrell PT, Marquinez NS, Correa JB, Meltzer LR, Unrod M, Sutton SK, . . . Brandon TH (2014). Expectancies for cigarettes, e-cigarettes, and nicotine replacement therapies among e-cigarette users (aka vapers). Nicotine & Tobacco Research, 17(2), 193–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. IBM Corp. (2017). IBM SPSS statistics for Windows, Version 25.0. Armonk, NY: IBM Corp. [Google Scholar]
  25. Janssen T, Treloar Padovano H, Merrill JE, & Jackson KM (2018). Developmental relations between alcohol expectancies and social norms in predicting alcohol onset. Developmental Psychology, 54(2), 281–292. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Jester JM, Wong MM, Cranford JA, Buu A, Fitzgerald HE, & Zucker RA (2015). Alcohol expectancies in childhood: Change with the onset of drinking and ability to predict adolescent drunkenness and binge drinking. Addiction, 110(1), 71–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kline RB (2015). Principles and practice of structural equation modeling: Guilford Publications. [Google Scholar]
  28. Leventhal AM, Strong DR, Kirkpatrick MG, Unger JB, Sussman S, Riggs NR, . . . Audrain-McGovern J (2015). Association of electronic cigarette use with initiation of combustible tobacco product smoking in early adolescence. JAMA, 314(7), 700–707. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Miller PM, Smith GT, & Goldman MS (1990). Emergence of alcohol expectancies in childhood: A possible critical period. Journal of Studies on Alcohol, 51(4), 343–349. [DOI] [PubMed] [Google Scholar]
  30. Montes KS, Witkiewitz K, Andersson C, Fossos N, Pace T, Berglund M, & Larimer ME (2017). Trajectories of positive alcohol expectancies and drinking: An examination of young adults in the US and Sweden. Addictive Behaviors, 73, 74–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Muthén L, & Muthén B (2014). Mplus (Version 7.3)[computer software].(1998–2014). Los Angeles, CA: Muthén & Muthén. [Google Scholar]
  32. Nelson SE, Van Ryzin MJ, & Dishion TJ (2015). Alcohol, marijuana, and tobacco use trajectories from age 12 to 24 years: Demographic correlates and young adult substance use problems. Development and Psychopathology, 27(1), 253–277. [DOI] [PubMed] [Google Scholar]
  33. Nicolai J, Moshagen M, & Demmel R (2012). Patterns of alcohol expectancies and alcohol use across age and gender. Drug and Alcohol Dependence, 126(3), 347–353. [DOI] [PubMed] [Google Scholar]
  34. Osberg TM, Billingsley K, Eggert M, & Insana M (2012). From animal house to old school: A multiple mediation analysis of the association between college drinking movie exposure and freshman drinking and its consequences. Addictive Behaviors, 37(8), 922–930. [DOI] [PubMed] [Google Scholar]
  35. Patrick ME, Wightman P, Schoeni RF, & Schulenberg JE (2012). Socioeconomic status and substance use among young adults: a comparison across constructs and drugs. Journal of Studies on Alcohol and Drugs, 73(5), 772–782. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Rohsenow DJ, Abrams DB, Monti PM, Colby SM, Martin R, & Niaura RS (2003). The smoking effects questionnaire for adult populations: Development and psychometric properties. Addictive Behaviors, 28(7), 1257–1270. [DOI] [PubMed] [Google Scholar]
  37. Rotter JB, Chance JE, & Phares EJ (1972). Applications of a social learning theory of personality. New York, NY: Holt, Rinehart Winston. [Google Scholar]
  38. Savolainen J, Eisman A, Mason WA, Schwartz JA, Miettunen J, & Järvelin M-R (2018). Socioeconomic disadvantage and psychological deficits: Pathways from early cumulative risk to late-adolescent criminal conviction. Journal of Adolescence, 65, 16–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Stueve A, & O’Donnell LN (2005). Early alcohol initiation and subsequent sexual and alcohol risk behaviors among urban youths. American Journal of Public Health, 95(5), 887–893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Trudeau L, Spoth R, Lillehoj C, Redmond C, & Wickrama K (2003). Effects of a preventive intervention on adolescent substance use initiation, expectancies, and refusal intentions. Prevention Science, 4(2), 109–122. [DOI] [PubMed] [Google Scholar]
  41. Volkow ND, Baler RD, Compton WM, & Weiss SR (2014). Adverse health effects of marijuana use. New England Journal of Medicine, 370(23), 2219–2227. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Wahl SK, Turner LR, Mermelstein RJ, & Flay BR (2005). Adolescents' smoking expectancies: Psychometric properties and prediction of behavior change. Nicotine & Tobacco Research, 7(4), 613–623. [DOI] [PubMed] [Google Scholar]
  43. Wood MD, Capone C, Laforge R, Erickson DJ, & Brand NH (2007). Brief motivational intervention and alcohol expectancy challenge with heavy drinking college students: A randomized factorial study. Addictive Behaviors, 32(11), 2509–2528. [DOI] [PubMed] [Google Scholar]
  44. Zamboanga BL, Horton NJ, Leitkowski LK, & Wang SC (2006). Do good things come to those who drink? A longitudinal investigation of drinking expectancies and hazardous alcohol use in female college athletes. Journal of Adolescent Health, 39(2), 229–236. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Table
NIHMS990014-supplement-Table.docx (13.5KB, docx)

RESOURCES