Revealed Preferences under Stochastic Attention: Theory and Experiment
GREDEG, Université Côte d'Azur (UCA), in Nice.
This paper investigates the empirical implementation of revealed preferences theory under stochastic limited attention. Because they are limited in their attention, individuals may failed to consider all the available alternatives and to choose the preferred one. It ultimately results in violation of the axioms usually needed in revealed preferences theory (WARP in deterministic choice and the regularity axiom in stochastic choice).
Recently, Brady and Rehbeck (2016, Econometrica) have proposed a set of axioms that uniquely characterize a “random conditional choice set rule”
(RCCSR). A RCCSR is a stochastic choice function such that the decision maker (DM) chooses its preferred alternative over a stochastic subset of the available alternatives (its consideration set). Moreover, if choice probabilities have a RCCSR representation, then it is possible to disentangle the DM’s preferences and consideration set. While being theoretically appealing, there has been no thorough empirical test of RCCSR.
Indeed, in this paper, we highlight statistical issues that hinder the empirical testability of the RCCSR, in presence of finite amount of choice data. We therefore propose a weak generalization of the RCCSR and provide new characterization and revealed preferences theorems, more adapted for empirical implementation. Based on this new characterization, we develop statistical procedures and investigate their type I and type II risks using numerical simulations. We ultimately confront these statistical procedures to a simple laboratory experiment, where preferences and limited attention are induced in a perceptual decision task. Our results show that subjects’ behaviors are consistent with weak RCCSR and we are able to reveal the preferences for most of the subjects.
Université Montpellier - Faculté d'économie, salle 416
Avenue Raymond Dugrand 34960 Montpellier
Dates & time