Background and aims:

Risk is inherent in human decision-making. Efficient measurement of domain-general risk preferences is crucial for large-scale surveys. This study examines how a widely used single-item measure of general risk preferences performs compared to established multi-item scales in predicting risk behaviour.

Methods:

Study 1 used a student sample (n = 486, Switzerland) and compared a single-item risk question (Dohmen et al. 2011) with a 10-item scale adapted from Dahlbäck (1990) against behaviour in an incentivized lottery choice experiment. Study 2 employed a nationally representative Swiss adult sample (n = 1’826) and compared the same single-item measure with the 8-item General Risk Propensity Scale (GRiPS; Zhang et al. 2019) in predicting choices in two hypothetical lottery tasks. Predictive validity was examined using OLS and linear probability models with demographic controls.

Results:

In study 1, the single-item measure predicted incentivized lottery behaviour more strongly than the multi-item scale. In study 2, both the single-item and GRiPS showed virtually identical predictive power: a 1 sd increase in either measure raised the probability of choosing the risky option by 12 %-points. Multi-item scales showed high internal consistency, yet no advantage in external validity. Known patterns – higher male risk tolerance and declining risk tolerance with age – were replicated across measures.

Discussion and conclusion:

A single general risk question provides predictive validity comparable to multi-item scales while minimizing respondent burden. For domain-general risk preferences, high measurement efficiency can be achieved without loss of validity, making the single-item instrument well suited for large surveys.