Individual Choice from a Convex Lottery Set: Experimental Evidence

This article reports on simple individual choice experiments, in which subjects choose a lottery from a convex set involving dominated lotteries and perfect negative correlation between risky assets. All subjects’ choices obey to [violate] the dominance criterion when dominance is [not] transparent. Choices involve too much risk since subjects overlook covariances. The main result is that elicited preferences suggest a two-step boundedly rational diversification pattern: subjects allocate a share of their endowment risklessly and allocate the remainder extremely risky but losslessly. This pattern is in line with safety first theory and loss aversion theory, but contradicts both in choice.