Sequential Search with Multiple Attributes: Optimal and Human Behaviour

Speaker: Prof Adam Sanjurjo

Affiliation: Universidad de Alicante

Zoom: https://uqz.zoom.us/j/82603079317

Abstract: Search theory is one of the pillars of economic theory, yet little is known about optimal sequential search in canonical multiattribute problems, where the complexity of the decision environment has traditionally rendered analytical and computational approaches intractable. We numerically characterize the optimal sequential search policy in environments with up to five alternatives and four attributes per alternative, where attributes can be searched separately, with full recall and unrestricted search order. We then compare human behavior to this benchmark in a between-subjects experiment that systematically varies problem dimensionality, search costs, and attribute dispersion. Subjects systematically deviate from the optimal policy along several dimensions, including total search intensity, responsiveness to search incentives, and the allocation of search across alternatives. The evidence suggests that subjects behave as if they compress the effective dimensionality of the search problem, concentrating search within a reduced subset of alternatives and attributes. A decision aid partially attenuates these distortions but leaves substantial performance gaps relative to the optimal benchmark.