Consumer Search : Not Enough or Too Much?

We study search behavior in a generalized "secretary problem" environment in which consumers search sequentially for the best alternative from a known and finite set of multi-attribute alternatives. In contrast to most previous studies, we make no distributional assumptions about the quality of the alternatives. Rather, at each stage of the search the consumers are only assumed to be able to rank order the alternatives they have already inspected in terms of their overall quality. Our study departs from previous experimental investigations of the secretary problem by including search costs and allowing for recall (backward solicitation) of previously inspected alternatives. Both the number of alternatives and the cost of searching are manipulated experimentally in a factorial design. In the no-cost condition, we find that subjects do not search enough, whereas in the cost condition they search too much. We propose a simple behavioral decision model that incorporates both local and global patterns of the sequence - patterns that should have been ignored by a rational consumer - and then show that it can account for the major patterns of the observed results