Assessing the construct validity of risk attitude

Two major approaches to measuring risk attitude are compared. One, based on the expected I utility model is derived from responses to lotteries and direct scaling. The other measure is a psychometric approach based on Likert statements that produces a unidimensional risk attitude scale. The data are from computer-assisted interviews of 346 Dutch owner-managers of hog farms, who made decisions about their own businesses. While the measures demonstrate some degree of convergent validity, those measures based on lotteries were better predictors of actual market behavior. In contrast the psychometric scale showed more agreement to self-reported measures of innovativeness, market orientation, and the intention to reduce risk. In light of the higher predictive validity of lottery-based measurements, we recommend elicitation methods based on the expected utility paradigm.