Integration of Limited Attention in Economic Theory

Overview & Synopsis

This thesis contributes to the integration of limited attention within the economic theory. We argue that attentional allocation processes can be understood as a production process with the allocated attention (the quantity of attentional resources invested in a decision) as an input and the effective attention (the amount of information contained in that decision) as an output. Borrowing methods from psychology and cognitive sciences, we propose three essays to shedding light on these processes. In the first chapter, we manipulate the presentation order between reward information and perceptual evidence in a two-alternative forced-choice task. The allocated attention is controlled, and we measure effective attention with a Signal Detection model. We found that the last information presented is more weighted in the decision. We attribute this effect to the division of attention. The second chapter proposes an experiment where participants pay costly attention to reduce the uncertainty of a discrimination task. We measure both allocated attention (through the response time) and effective attention (through performance). This experiment allows the study of attentional social dilemmas (situations where attention is costly for individuals but beneficial for the group). We highlight a discrepancy between monetary elicited social preferences and the behaviors exhibited in our attentional social dilemma. The last chapter proves that a model of revealed preferences under stochastic attention can be implemented and tested empirically. We provide new characterization and revealed preference theorems for a general version of Brady and Rehbeck’s model (2016, Econometrica). We propose and analyze – with numerical simulations – statistic procedures to test the axioms, to reveal preferences, and to measure effective attention. We test the internal validity of the model with a selective attention task, where participants choose an alternative among distractors and we find that most of the participants behave in accordance with the model and reveal coherent preferences.