Information-Constrained Discrete Choice ∗

It is not unusual in real-life that one has to choose among finitely many alternatives when the merits of each alternative are not perfectly known. A natural rule is then to choose the alternative which “looks” best. We show that a version of the monotone likelihood ratio property is sufficient, and also essentially necessary, for this decision rule to be optimal. We also analyze how the precision of the decisionmaker’s information affects his or her choice and welfare, and we show that it is not always advantageous to have more choice alternatives. We extend the analysis to situations in which the decision-maker at some cost or effort can choose the precision in his or her information, as well as the number of alternatives to consider. We show that small differences in information cost can result in dramatic differences in choice behaviors. Individuals for whom information gathering is costly will optimally make ill-informed choices. We also show that welfare is not always monotonic in true utilities, both under exogenous and endogenous uncertainty.