International audienceWe present a new experimental evidence of how framing affects decisions in the context of a lottery choice experiment for measuring risk aversion. We investigate framing effects by replicating the Holt and Laury's (Am. Econ. Rev. 92:1644-1655, 2002) procedure for measuring risk aversion under various frames. We first examine treatments where participants are confronted with the 10 decisions to be made either simultaneously or sequentially. The second treatment variable is the order of appearance of the ten lottery pairs. Probabilities of winning are ranked either in increasing, decreasing, or in random order. Lastly, payoffs were increased by a factor of ten in additional treatments. The rate of inconsistencies was significantly higher in sequential than in simultaneous treatment, in increasing and random than in decreasing treatment. Both experience and salient incentives induce a dramatic decrease in inconsistent behaviors. On the other hand, risk aversion was significantly higher in sequential than in simultaneous treatment, in decreasing and random than in increasing treatment, in high than in low payoff condition. These findings suggest that subjects use available information which has no value for normative theories, like throwing a glance at the whole connected set of pairwise choices before making each decision in a connected set of lottery pairs

P reference reversals have been widely studied using risky or riskless gambles. However, little is known about preference reversals under ambiguity (unknown probabilities). Subjects were asked to make a binary choice between ambiguous P-bets (big likelihood of giving small prize) and ambiguous $-bets (small likelihood of giving large prize) and their willingness to accept was elicited. Subjects then performed the same two tasks with risky bets, where the probability of winning for a given risky bet is the center of the probability interval of the corresponding ambiguous bet. Preference reversals are not only replicated under ambiguity but are even stronger than are those under risk. This is due to higher elicited prices for the $-bet and lower elicited prices for the P-bet under ambiguity than under risk. This result can be explained by the shape of the weighting function for different levels of uncertainty and for different elicitation modes.