Correlation Neglect in Student-to-School Matching

A growing body of evidence suggests that many people struggle with decision-making in the presence of correlation. In typical examples of this problem, decision-makers are presented with multiple signals that are each influenced both by independent components and information from a common source. The process by which signals are generated induces correlation, and optimal decision-making requires taking it into account. In practice, however, experiments like those of Enke and Zimmermann [1] demonstrate that many decision-makers neglect to do so, effectively acting as if these correlated signals are independent. We study the prevalence and consequences of these failures of reasoning in a decision of considerable importance: the application strategies of students applying to schools. Many application processes inherently require students to make forecasts of events determined by common underlying inputs, resulting in correlation structures like those described above. For example, students commonly must whittle a large number of schools down to a smaller set that are applied to or ranked, introducing an incentive to avoid listing two programs with highly correlated admissions decisions. In such environments, a student harboring correlation neglect faces a challenging decision.