Random network consideration: Theory and experiment

In many settings, it is natural to think of limited consideration exhibiting spillovers: attention paid to a particular alternative may “spill over” to another alternative based on shared characteristics, complementarities, features of the choice environment, shared advertising campaigns, product bundling, etc. Limited consideration of this form gives rise to new methods of revealing preferences and attention. Using a novel laboratory experiment, I test the attention properties of a deterministic Network Consideration model proposed in previous work and find a plethora of violations thereof, even at the individual level. I then propose a stochastic generalization, Random Network Consideration, and analyze its properties regarding the formation of consideration sets. When applied to the laboratory data, I find greater consistency with the general Random Network Consideration model. These results reveal that intuitively appealing features of theories of limited attention may indeed be too stringent in practice, an insight which should guide future theoretical research.