Eliciting risk preferences: is a single item enough?

Abstract

Economists and psychologists frequently use single-item measures of risk preferences despite potential limitations in reliability and criterion validity compared to their multi-item counterparts. This can be particularly problematic when individual differences in risk preferences are used to predict real-world economic, health, and financial outcomes. In this paper, we compare a popular single-item measure of risk preference, the General Risk Question (GRQ), to multi-item measures of domain-general and -specific risk preference measures. In a two-wave survey study of 434 adults, we found that the GRQ had good psychometric reliability and converged with other multi-item measures of risk preferences. The GRQ also exhibited a similar pattern of associations with other personality and demographic variables as compared to multiitem measures. However, the predictive validity of the GRQ was lower than multi-item measures for most of the outcomes examined. The GRQ also explained less incremental variance for realworld outcomes over the Big Five personality traits than the multi-item counterparts. Although the GRQ is a construct-valid measure of risk preferences, researchers should nonetheless consider the trade-off between survey efficiency and predictive efficacy when deciding whether a single item is enough.