Body mass predicts personality development across 18 years in middle to older adulthood

Abstract

Objective: Various personality traits have longitudinal relations with body mass index (BMI), a measure of body weight and a risk factor for numerous health concerns. We tested these associations' compatibility with causality in either direction. Method: Using three waves of the Wisconsin Longitudinal Study (N = 12,235, M age = 53.33 at baseline), we tested how accurately the Five-Factor Model personality domains and their items could collectively predict BMI and change in it with elastic net models. With multilevel models, we tested (a) bidirectional and (b) within-person associations between BMI and personality traits. Results: The five domains were able to predict concurrent (r = 0.08), but not future BMI. Twenty-nine personality items predicted concurrent and future BMI at r = 0.21 and r = 0.16 to 0.25, respectively. Neither the domains nor items could collectively predict change in BMI. Similarly, no individual trait predicted change in BMI, but BMI predicted changes in Conscientiousness, Agreeableness, and several items (|b*| = 0.03 to 0.08). BMI had within-person correlations with these same traits; time-invariant third factors like genetics or childhood environments therefore could not (fully) account for their relations. Conclusions: Body weight may contribute to adults' personality development, but the reverse appears less likely.

A bottom-up approach dramatically increases the predictability of body mass from personality traits

“…To provide some context for this result, several studies have used personality items in similar machine learning models to predict BMI with accuracies generally not exceeding r = .25 [ 10 , 31 , 32 ], but being r = .43 at maximum when using a pool of 135 items [ 27 ]. Relevant to the current results, the latter study also tested prediction at different imposed levels of data missingness, finding prediction with complete personality data to be 269% higher than at 90% missingness.…”

“…Fourth, high data missingness limited statistical power in calculating the factors’ correlations with BMI and likely affected predictive accuracy as predictions are considerably more accurate with complete data [ 27 ]. Fifth, the data were cross-sectional and did not enable conclusions about possible longitudinal associations—but given that longitudinal associations between personality traits and BMI are roughly similar to cross-sectional ones [ 10 ], it is likely the same traits could predict future BMI with similar accuracy as they can concurrent BMI. And finally, although the sample was geographically diverse, the analyses were not focused on cultural differences, but different traits may be linked to body weight between cultures [ 39 ] and the results may not generalize to all populations.…”

“…Body weight matters for life quality in several ways: Not only does excessive weight predict adverse health outcomes [ 7 ] and lower subjective well-being [ 8 ] but it is also associated with stigmatization and discrimination in many life domains, from education and employment to interpersonal relationships [ 9 ]. On one hand, body weight can be predicted from sets of personality traits [ 10 ], which could have applications in health care. On the other hand, its descriptive correlations with individual personality traits and trait groups can inform hypotheses regarding the causes and consequences of body weight.…”

“…The current study relies on cross-sectional data and by predicting BMI we mean applying predictive models trained in one sample to other, independent samples. However, there is some evidence that personality items have similar relations to concurrent and future BMI [ 10 ], suggesting that the same traits that predict BMI concurrently also predict its future values (perhaps as a result of traits’ effects on weight accumulating over time). As personality traits could have useful clinical applications—for instance, they could be used to stratify people based on their risk for adverse outcomes—development of brief but predictively accurate assessment methods may be particularly useful [ 18 ], especially for such a common condition like overweight with a range of adverse sequelae.…”