Abstract.
We consider a model of sequential search in which a decision-maker (DM) has to choose one alternative from a fixed set. All available alternatives are iid random variables and ex-ante unknown to the DM. Before making a choice, contrary to the standard search literature, we allow the DM to decide how much and what kind of information to acquire about each alternative, e.g., design different job market interviews for candidates with different arrival ranks. We find that optimal interviews have an intuitive property – the first arriving candidates are treated harshly, and their interviews are harder to pass, while later candidates’ interviews are easier to pass. We compare the unconditional probabilities of choice and study the discrimination the order of inspection can cause. We argue that discrimination is sensitive to the functional form of the cost of learning. We consider several extensions, and we show that it may be optimal for the DM to interview an inferior candidate first and that a naive affirmative action policy can increase discrimination.
