Behavioral economics and the aggregate versus proximal impact of sociality on heavy drinking

Abstract

Background: Behavioral economic theory predicts decisions to drink are cost benefit analyses, and heavy episodic drinking occurs when benefits outweigh costs. Social interaction is a known benefit associated with alcohol use. Although heavy drinking is typically considered more likely during more social drinking events, people who drink heavily in isolation tend to report greater severity of use. This study explicitly disaggregates between-person and within-person effects of sociality on heavy episodic drinking and examines behavioral economic moderators.

Methods: We used day-level survey data over an 18-week period in a community adult sample recruited through crowdsourcing (mTurk; N = 223). Behavioral economic indices were examined to determine if macro person-level variables (alcohol demand, delay discounting, proportionate alcohol-related reinforcement [R-ratio]) interact with event-level social context to predict heavy drinking episodes.

Results: Mixed effect models indicated significant between-person and within-person social context associations. Specifically, people with a higher proportion of total drinking occasions in social contexts had decreased odds of heavy drinking, whereas being in a social context for a specific drinking occasion was associated with increased odds of heavy drinking. Person-level R-Ratio, demand elasticity, and breakpoint variables interacted with social context to predict heavy episodic drinking, such that the event-level social context association was stronger when R-Ratios, alcohol price insensitivity, and demand breakpoints were high.

Conclusions: These results demonstrate an ecological fallacy, in which the size and direction of effects were divergent at different levels of analysis, and highlight the potential for merging behavioral economic variables with proximal contextual effects to predict heavy drinking.

Keywords: Alcohol; Behavioral economics; Decision-making; Demand; Discounting; Social.

Figure 1.. Predicted Probabilities for Behavioral Economic Interaction Models.

Plotted are the predicted probabilities from generalized linear mixed effect models predicting heavy drinking days on days of alcohol use. Values are presented across the full range of the mean-centered behavioral economic variable. Location of the area including plus one or minus one standard deviation from the mean is provided by the vertically dashed lines. Predictions are plotted assuming the mean for the between-subject social context variable. Note that a linear transformation of demand elasticity values was conducted (x100) to facilitate model convergence.