Strategic Sample Selection

This paper develops a notion of multivariate accuracy to characterize the welfare impact of sample selection from a larger presample. Maximal selection benefits or hurts a decision maker with interval dominance ordered preferences if the reverse hazard rate of the data distribution is log-supermodular—as in location experiments with normal noise—or log-submodular. Applying the result to auctions, we show that under non-pathological conditions the information contained in the winning bids decreases as the number of bidders increases. Exploiting a connection to extreme value theory, we quantify the limit amount of information when the presample size goes to infinity, as under perfect competition. In the context of a model of equilibrium persuasion with costly information, we also derive implications for the design of selected experiments when selection is made by an examinee, a biased researcher, or contending sides with the peremptory challenge right to eliminate a number of jurors.