Xinguang Chen, University of Florida, Gainesville, FL and Wuhan University, Wuhan, China
Yan Wang, University of Florida, Gainesville, FL
Ding-Geng Chen, University of North Carolina, Chapel Hill, NC

Abstract: Findings from cusp catastrophe modeling of binge drinking among college students in the United States (U.S.) and high school students in other countries (i.e., Japan) have advanced our understanding of alcohol abuse. No reported study has used the same method to investigate binge drinking among U.S. high school students, a high-risk population for alcohol abuse. In this study, we analyzed data for 12th graders (n = 1122) from the 2015 Monitoring the Future (MFT) Study with cusp catastrophe models. Guided by Kahneman s dual-system theory that integrates the theories of reasoned action and planned behavior, perceived alcohol harm (0-3, larger indicating less harm) was modeled as asymmetry variable (knowledge and beliefs), and drinking in social settings (contextual factors, 0 = no and 4 = always) as bifurcation variable to predict binge drinking in the past two weeks (0, 1, 2, 3 or more times). Analytical results indicated that a cusp model, depending on the method for parameter estimation could explain 36-88% of the variance in binge drinking while a linear model only explained up to 14%. Results from the indirect cusp modeling methods indicated that perceived alcohol harm and drinking in social settings both significantly predicted binge drinking. However, the estimated cusp point varied depending on the method for parameter estimation. In conclusion, the dynamics of binge drinking among U.S. high school students are nonlinear and discrete. Knowledge-based interventions to prevent binge drinking need to emphasize social and contextual factors that may trigger sudden behavior change to achieve greater intervention effect. To advance cusp catastrophe modeling, further research is needed to improve the method for parameter estimation.

Keywords: cusp catastrophe modeling, binge drinking, adolescents, nonlinear discrete system