Successful Explanations Start with Accurate Descriptions: Questionnaire Items as Personality Markers for More Accurate Predictions

Abstract

Personality–outcome associations, typically represented using the Big Five personality domains, are ubiquitous, but often weak and possibly driven by the constituents of these domains. We hypothesized that representing the associations using personality questionnaire items (as markers for personality nuances) could increase prediction strength. Using the National Child Development Study (N = 8719), we predicted 40 diverse outcomes from both the Big Five domains and their 50 items. Models were trained (using penalized regression) and applied for prediction in independent sample partitions (with 100 permutations). Item models tended to out–predict Big Five models (explaining on average 30% more variance), regardless of outcomes’ independently rated breadth versus behavioural specificity. Moreover, the predictive power of Big Five domains per se was at least partly inflated by the unique variance of their constituent items, especially for generally more predictable outcomes. Removing the Big Five variance from items marginally reduced their predictive power. These findings are consistent with the possibility that the associations of personality with outcomes often pertain to (potentially large numbers of) specific behavioural, cognitive, affective, and motivational characteristics represented by single questionnaire items rather than to the broader (underlying) traits that these items are ostensibly indicators of. This may also have implications for personality–based interventions. Copyright © 2018 European Association of Personality Psychology