Your personality data can do more: Items provide leverage for explaining the variance and co-variance of life outcomes

Abstract

Drawing parallels with genetics, we propose that representing associations of personality with life outcomes using individual questionnaire items can provide a low cost leverage of existing knowledge base and data. For illustration, we show that item-based models, trained in sample to predict eleven outcomes, clearly outperformed models based on Five-Factor Model (FFM) domains or facets in an independent sample, with average proportions of explained variance being 8% (for the item-based model), 3.4% (domain-based) and 5.7% (facet-based). This suggests that personality-outcome associations are often driven by specific characteristics represented by single items (sometimes called nuances) rather than higher-order structures purported to underlie the items. Item-based models also tend to have the highest predictive specificity (discriminant validity), whereas the FFM domains are most likely to predict outcomes' general valence rather than their distinctive aspects. Similar to polygenic scores widely used in genetics, item-level effects can be aggregated into polyitem scores, whereas correlations among polyitem scores can be interpreted as personality correlations (analogous to genetic correlations). Personality correlations quantify the extent to which outcomes overlap in personality correlates and can help to understand how personality is correlated with outcomes or account for co-variations of outcomes (e.g., by calculating the proportion of the correlation between outcomes that can be ascribed to personality characteristics). An empirical illustration focuses on body mass index, potentially related lifestyle aspects, and educational level. Overall, the proposed approach allows gleaning more and conceptually novel information about how personality may intersect with life outcomes—from already existing data.