Top Down or Bottom Up? Disentangling the Channels of Attention in Risky Choice

Economists have become increasingly interested in using attention to explain behavioral patterns both on the micro and macro level. This has resulted in several disparate theoretical approaches. Some, like rational inattention, assume a "top-down" model of executive optimization. Others, like salience theory, assume a "bottom-up" influence where attention is driven by contextual factors. This distinction is fundamental for the economic implications of attention, but so far there is little understanding of their relative importance. We propose a multi-attribute random utility model that unifies prior theoretical approaches by distinguishing between the impact of top-down and bottom-up attention. We accomplish this by separating agent-specific and decision-specific variation in attention and verify our framework in an eye-tracking experiment on risky choice. We find that both top-down and bottom-up attention are connected to important choice variables: both are associated with the weighting of the attributes of choice options, while top-down attention is additionally associated with measures of loss aversion. We discuss the insights regarding the nature of attention and its role in economic theory.