This paper evaluates risk preference measures by contrasting subjective or self-assessed risk with objective risk, as implicated by bank customers’ actual portfolio allocation. Using a detailed data set of 7,234 bank customers, we find that subjective risk measures can explain and predict objective risk, but that the relationship is relatively weak. Subjective measures that uses survey questions about the customers’ trade-off between risk and return is a better measure than the hypothetical lottery for explaining objective risk. Both measures are relatively weak at predicting objective risk, but perform better than using a naïve model. We also find that multiple-item variables are somewhat better than single-item variables for explaining objective risk.