Optimal Sequential Search Among Alternatives

“Optimal Sequential Search Among Alternatives”with Michael Choi

We explore costly sequential search among finitely many risky options, and an outside option. Payoffs are the sum of a known and hidden random factor.
(a) We resolve a long open question about how riskier payoffs impact search duration: expected search time is higher for more dispersed idiosyncratic noise.
(b) Since options differ ex ante, we incorporate selection effects into search: Counterintuitively, with few options, the quitting chance falls if search costs rise; also, while stopping rates rise over time, earlier options are recalled more.
(c) We find that the stationary search model is a misleading benchmark: For as the number of options explodes, the recall chance is bounded away from zero if the known factor has a distribution without a thin tail (eg. exponential).
(d) A special case of our model captures web search engines that rank order options: We prove that the “click through rate” (the chance of initiating a search) is a poor quality measure since it falls in accuracy for expensive goods.