Abstract
Objective: Adolescence is characterized by psychosocial and cognitive changes that can alter the perceived risk of negative effects of alcohol, opportunities to drink, and self-control. Few studies have investigated whether these factors change in their contribution to adolescent drinking over time. This study examined associations between perceived risk, opportunities to drink, self-control, and past-year drinking and investigated whether self-control buffers the effect of lower perceived risk and frequent drinking opportunities on the probability of past-year drinking. Method: Data from a four-wave longitudinal study (2015–2020) of 2,104 North Carolina adolescents (Mage = 12.36, SDage = 1.12, at Time 1) were used to assess changes in associations between self-control, perceived risk of drinking, and drinking opportunities on the frequency of past-year drinking. Hypotheses were tested using latent trajectory models. Results: At all timepoints, greater perceived risk, fewer drinking opportunities, and higher self-control were associated with drinking abstinence in the past year. Self-control buffered the impact of frequent drinking opportunities and lower perceived risk on the probability of alcohol use at Times 1–3. Conclusions: Despite expectations that adolescents’ ability to navigate their environments improves as they age, associations between risk, protective factors, and past-year drinking were relatively stable over time. Nevertheless, self-control protected against frequent drinking opportunities and lower perceived risk. Strategies that support or relieve the need for self-control (e.g., situation modification) may protect against alcohol use throughout adolescence.
Acknowledgments
We are grateful to Erin Davisson, Anastasia Georgiades, Ann Skinner, and Sarah Kwiatek for their valuable feedback at various stages of this work.
Disclosure statement
The authors report no conflicts of interest.
Notes
1 Main effects hypotheses were pre-registered at https://osf.io/xqanc; anticipated results of moderating effects of self-control were not pre-registered.
2 In light of limitations of alpha coefficients (McNeish, Citation2018), we report omega coefficients of internal consistency, which are less positively biased than alpha coefficients (Flora, Citation2020).
3 The original pre-registered model could not be estimated due non-convergence despite extensive troubleshooting. We thus tested our research questions using a back-up approach, as outlined in the pre-registration in the event of non-convergence. The approach allowed the introduction of interaction effects between the variables of interest, which were not possible in the original model.
4 Although these differences suggest that data was not missing at random, the inclusion of these covariates in all models has the effect of making them so.
5 χ2(15) = 21.41, p = .124; RMSEA = 0.01; CFI = 0.99, TLI = 0.98; SRMR = 0.04.
6 χ2(51) = 84.47, p = .002; RMSEA = 0.02; CFI = 0.99, TLI = 0.96; SRMR = 0.03.
7 χ2(51) = 69.65, p = .042; RMSEA = 0.01; CFI = 0.99, TLI = 0.97; SRMR = 0.03.
8 χ2(51) = 121.67, p < .001; RMSEA = 0.03; CFI = 0.98, TLI = 0.92; SRMR = 0.03.