Knowing more and Step by step: Correlation neglect under preference uncertainty in centralized school choices

Abstract

We conduct a laboratory school-choice experiment in which admission cut-offs arise endoge- nously, testing correlation neglect in school choice settings. Relative to Rees-Jones et al. (2024) which fixes thresholds ex ante, we show that correlation neglect persists in complex school-choice environments. We evaluate three debiasing tools. First, we provide a reminder highlighting that admission outcomes are correlated. Second, we offer an algorithmic aid that supplies personal- ized admission probabilities, reducing the burden of Bayesian updating. Third, we implement an iterative deferred-acceptance mechanism that reveals rejection outcomes sequentially, sim- plifying contingent reasoning. The reminder has no effect. Providing personalized probabilities substantially reduces aggressive applications, but the reduction does not vary with the severity of correlation neglect. In contrast, the iterative mechanism directly attenuates correlation-neglect errors, especially when correlated risks are large. These findings inform the ongoing global reform of centralized admissions systems and suggest that both informational and structural interventions deserve policy attention.