Substitutions probabilistiques et décision individuelle devant le risque. Enseignements de l'expérimentation

This paper uses laboratory experiments as tools allowing to weigh comparatively some of the new models of decision making under risk which have been proposed in the last twenty years to describe behavior towards risk. These models rest on diverging interpretations of the observed gaps with respect to the expected utility model. The interpretation put forward by the "rank dependent" model ? independently suggested by J. Quiggin, M. Yaari and M. Allais ? can be shown to be based on the concept of "probabilistic substitution". This interpretation is reviewed and contrasted with the one provided by the models which preserve the betweenness property of indifference loci. Instead of trying to fit the different models with existing data from observation, experiments try to look at subjects? behavior as revealed through the expression of indifferences and to compare such observations with the implications of the models under scrutiny. This research programme, developed at GRID during several years and briefly sketched here in few pages, leads to conclude that the descriptive ability of the rank dependent model appears as the most satisfactory.