Are preferences incomplete? An experimental study using flexible choices

Completeness, the most commonly assumed axiom in preference theory, has not received much attention from the experimental literature. Indeed, incomplete preferences model a cognitive phenomenon (an agent's inability to compare alternatives), and therefore cannot be directly revealed through choice behavior. Implementing a solution to this methodological issue recently proposed by Danan [A behavioral model of individual welfare, mimeo EUREQua University Paris 1, 2003], we build an experimental protocol involving choices among menus of lotteries, and reveal cognitive preferences' incompleteness by means of the concept of preference for flexibility. Our experimental protocol is designed to assess the descriptive validity of the completeness axiom, as well as to relate its possible violations to lotteries' riskiness. Two-thirds of the subjects whose choices reveal preferences in accordance with the underlying theory exhibit a strictly positive measure of incompleteness. The observed average measure of incompleteness equals approximately 17 percent and it is significantly greater than 10 percent. We do not find a significant relationship between a lottery's riskiness and its cognitive comparability with certain payoffs.