Delay discounting of rewards and losses, alcohol use, and the influence of socioeconomic factors: A cross‐sectional online study in frequent drinkers

Mathieu Pinger, Malin Skirke, Janine Thome, Wolfgang H. Sommer, Georgia Koppe, Peter KirschAlcohol 2024 · DOI

TL;DR

This study investigated how "delay discounting" (valuing immediate rewards over future ones) relates to alcohol use, and whether socioeconomic factors like income and education play a confounding role. The findings show a small but consistent link between steeper reward discounting and higher alcohol consumption, but this association is weakened by education and age, suggesting these factors are important to consider in future research.

What Problem Does This Paper Address?

The paper tackles the complex relationship between delay discounting and problematic alcohol use. Delay discounting is a behavioral economics concept that describes how people devalue future outcomes compared to immediate ones. In addiction research, steeper delay discounting (i.e., a stronger preference for immediate rewards) has been consistently linked to problematic alcohol use. However, a critical gap exists: most studies haven't adequately accounted for socioeconomic status (SES), such as income and education, which are also known to influence both delay discounting and alcohol consumption. This omission raises questions about whether the observed link between delay discounting and alcohol use is a direct predictor or a spurious correlation driven by SES. The study aims to clarify this by investigating the robustness of delay discounting as a predictor of alcohol use after controlling for SES. Additionally, it explores the less-studied area of delayed loss discounting (how people devalue future negative outcomes) in relation to alcohol use.

Methods

The researchers employed a cross-sectional online study design. Participants completed an intertemporal choice task to measure delay discounting of both rewards and losses. This task involved making decisions between hypothetical immediate and delayed monetary outcomes. The researchers used a hyperboloid discounting model to infer discounting parameters (specifically, the log-transformed discounting parameter, log(κ), for both rewards and losses) from participants' choices. This model is a standard approach in behavioral economics for quantifying delay discounting.

In addition to the discounting task, participants completed several self-report questionnaires:

  • Alcohol Use Disorder Identification Test (AUDIT): To assess problematic alcohol use.
  • Daily Drinking Questionnaire (DDQ): To measure weekly alcohol consumption.
  • Barratt Impulsiveness Scale (BIS-15): To measure impulsivity.
  • Various questionnaires assessing socioeconomic status, including objective measures (education, income) and subjective measures (perceived financial well-being).

Data analysis involved linear and hierarchical regression analyses to test the relationships between delay discounting, SES, and alcohol use, while controlling for confounding variables.

Datasets

The study utilized data collected from 341 participants recruited through the online platform Prolific. This sample consisted of moderate-to-heavy-drinking individuals residing in the UK. No new benchmark datasets were introduced; the study relied on existing measures and participant recruitment platforms.

Key Insights and Ideas

  • The "small but robust" association: The study confirms a small but statistically significant association between delay discounting of rewards (DRD) and problematic alcohol use. This aligns with a large body of prior research.
  • SES as a confounder: Crucially, the study demonstrates that education and age significantly influence the relationship between DRD and alcohol use. When these factors are controlled for, the predictive power of DRD diminishes, suggesting they act as confounders. This highlights the importance of accounting for SES in future research.
  • Income's weaker role: Interestingly, income did not show a significant confounding effect on the DRD-alcohol use association in this study, contrary to some prior findings. This might be due to the specific income categories used or the nature of the online participant pool.
  • DLD is not a strong predictor: Unlike DRD, delayed loss discounting (DLD) showed no significant association with problematic alcohol use in this study. This suggests that devaluing future negative outcomes might not be as relevant a factor in problematic alcohol use as devaluing future rewards.
  • Impulsivity vs. Discounting: The study found that while impulsivity (as measured by BIS-15) was correlated with both DRD and alcohol use, DRD did not explain unique variance in AUDIT scores when impulsivity was already accounted for. This supports the idea that DRD might not simply be a measure of impulsivity.

Key Results

  • DRD and AUDIT scores: A small but significant positive correlation was found between DRD (log(κR)) and AUDIT scores (r=0.15, p=0.01).
  • DRD and weekly alcohol consumption: A similar small but significant positive correlation was observed between DRD (log(κR)) and weekly alcohol consumption (r=0.12, p=0.01).
  • DLD and AUDIT scores: No significant correlation was found between DLD (log(κL)) and AUDIT scores (r=0.09, p=0.10).
  • SES and DRD: Education was negatively correlated with DRD (r=-0.19, p<0.01), meaning higher education was associated with less steep reward discounting.
  • SES and alcohol use: Both education (r=-0.15, p<0.01) and income (r=-0.12, p=0.03) were negatively correlated with AUDIT scores.
  • Confounding effect of SES: When controlling for education and income, DRD remained a significant predictor of AUDIT scores, but the explained variance was marginal (AR2=0.015, p=0.02). However, when age and education were included, DRD did not explain significant additional variance.

Limitations and Future Work

  • Cross-sectional design: The study's cross-sectional nature prevents the establishment of causal relationships. Future longitudinal studies are needed to confirm these findings and explore causal pathways.
  • Online sample: The use of an online platform (Prolific) may have introduced sampling biases, potentially undersampling individuals with lower SES or severe addiction. This limits the generalizability of the findings to clinical populations.
  • Income categories: The broad income categories used might have masked potential effects of income on DRD. Future studies could use more granular income measures.
  • Monetary discounting: The study used monetary outcomes for discounting. The authors suggest that monetary aversion discounting may not be a useful task in addiction research, as negative consequences of alcohol use are often psychological (e.g., craving, withdrawal) rather than purely monetary.
  • Confounding variables: The study highlights the need for future research to systematically account for socioeconomic and demographic variables when investigating DRD and alcohol use. The potential confounding effects of age and education warrant further investigation.
  • Larger samples: The authors suggest that larger samples are needed for reliable estimation of DRD's unique effects, especially when accounting for multiple covariates.

Summary updated

Abstract

We replicated a small but robust association between alcohol use and DRD, but not DLD. This association appeared to be confounded by education and age, but not by income. We conclude that socioeconomic and demographic variables should systematically be accounted for in future studies investigating DRD and alcohol use.
BibTeX
@inproceedings{pinger2024delay,
  author = {Mathieu Pinger and Malin Skirke and Janine Thome and Wolfgang H. Sommer and Georgia Koppe and Peter Kirsch},
  title = {{Delay discounting of rewards and losses, alcohol use, and the influence of socioeconomic factors: A cross‐sectional online study in frequent drinkers}},
  booktitle = {Alcohol},
  year = {2024},
  doi = {10.1111/acer.15469},
}
Figure