People’s risk preferences are thought to be central to many consequential real-life decisions, making it important to identify robust correlates of this construct. Various psychological theories have put forth a series of candidate correlates, yet the strength and robustness of their associations remain unclear because of disparate operationalizations of risk preference and analytic limitations in past research. We addressed these issues with a study involving several operationalizations of risk preference (all collected from each participant in a diverse sample of the German population; N = 916), and by adopting an exhaustive modeling approach—specification curve analysis. Our analyses of 6 candidate correlates (household income, sex, age, fluid intelligence, crystallized intelligence, years of education) suggest that sex and age have robust and consistent associations with risk preference, whereas the other candidate correlates show weaker and more (domain-) specific associations (except for crystallized intelligence, for which there were no robust associations). The results further demonstrate the important role of construct operationalization when assessing people’s risk preferences: Self-reported propensity measures picked up various associations with the proposed correlates, but (incentivized) behavioral measures largely failed to do so. In short, the associations between the 6 candidate correlates and risk preference depend mostly on how risk preference is measured, rather than whether and which control variables are included in the model specifications. The present findings inform several theories that have suggested candidate correlates of risk preference, and illustrate how personality research may profit from exhaustive modeling techniques to improve theory and measurement of essential constructs.