Temporal discounting does not influence body mass index

Abstract

The prevalence of obesity has driven searches for cognitive or behavioural economic factors related to Body Mass Index (BMI). One candidate is delay discounting: those who prefer smaller sooner rewards over larger but later rewards are hypothesised to have higher BMI. The findings in the literature are mixed however, with meta analyses suggesting only a very small correlation between discounting and BMI. Here we present novel empirical data (N = 381) and Bayesian analyses which suggest no such relationship between discounting of either monetary or weight loss rewards and BMI. We also find evidence against our novel proposal that discounting moderates the rate of BMI gain over time. We also present our data in the context of a random effects Bayesian meta-analytical result which does suggest the presence of a small correlation overall. The strength of the correlation is so weak (2.25% shared variance) that its practical significance may be minor to non existent. However because we found decisive evidence for unaccounted for study-level variance, due to study heterogeneity, we argue that we should treat such meta-analytic correlations with extreme caution. While the relationship between discounting and health outcomes such as BMI remain theoretically appealing, our empirical and meta-analytic results suggest we should be cautious in inferring a correlational, let alone a causal, role for discounting processes in driving BMI or moderating BMI gain with age.

No simple link between temporal discounting and body composition

“…There was also evidence against subjective hunger being correlated with discounting of money ( = 0.034, 10 = 0.081, = 359) or food ( = 0.086, 10 = 0.193, = 238). Having failed to find compelling evidence for correlations between discounting and body composition, we tested: a) if there was a main effect of discount rate after accounting for the effects of age on body composition, and b) whether discount rates moderate the rate of increase of BMI or WHtR (see the process model of Veillard & Vincent, 2020). We evaluated these hypotheses in a series of Bayesian multiple linear regressions, examining the relationship of: 1) BMI as a function of age and discount rate for money, 2) WHtR as a function of age and discount rate for food, 3) BMI as a function of age and discount rate for money, 4) WHtR as a function of age and discount rate for food.…”

“…We conducted 2 online studies -the first measured delay discounting for money and food, and the second measured delay discounting for weight loss rewards and food rewards (see Table 1). We re-analysed the data from Veillard and Vincent (2020), resulting in the same conclusions being drawn. Finally, we completed a meta analysis where we pooled data across studies to achieve even larger sample sizes.…”

“…Because of our recruitment methods we believe the majority of participants responding will have been from the UK, but we cannot exclude the possibility that some participants were from other countries. We cannot exclude the chance of some participant overlap between the current studies and Veillard and Vincent (2020), but it is unlikely as the data was collected one year apart.…”

“…Jarmolowicz et al, 2014). In summary, despite the arguments presented by Veillard and Vincent (2020), there are still compelling reasons to suspect a link between discounting and body composition. Therefore, we sought to further explore this possible relationship.…”

“…A recent effort at this was reported by Veillard and Vincent (2020). In a sample of 318 participants, Bayesian methods provided evidence against correlations between discounting of monetary or weight loss rewards and BMI.…”