Eliciting risk-preferences in socio-economic surveys: How do different measures perform?

•I compare three different measures of risk attitudes collected in a general survey.•External validity, data quality and predictive validity of the measures are compared.•Lottery choice tasks can be implemented in non-incentivized surveys.•Domain specific risk measures are more appropriate to predict behaviour.•Short form of popular psychometric questions work well in general surveys.

This paper contributes to the debate on the adequate elicitation of individual risk attitudes in general socio-economic surveys. A multi-item question on the willingness to take risk, a very short form of the DOSPERT scale (Weber et al., 2002) and a series of lottery tasks are compared with respect to the quality of the answers and the predictive validity of the derived risk measures. The quality of the collected data appears to be high. All the measures are informative about individual's attitudes while item nonresponse is mostly unproblematic. The measures however differ in their predictive power, with the lottery-based measures exhibiting only weak predictive validity. When the scope of the assessment is to predict behaviour, domain specific risk measures seem to be more appropriate. Embedding a short DOSPERT scale in general surveys appears to be very promising for empirical applications in social sciences that use survey-based risk measures.