A Bayesian Model of Attention Allocation: Applications in Gaokao School-Choice Reform

The prevailing assumption implicit in the seminal works of school-choice matching is that students have fully informed preferences. As a contribution to the literature, we consider how the information acquisition costs that students incur while learning
about their preferences factor into the welfare calculus. Specifically, we adopt cost-effectiveness as a desideratum of mechanism selection and introduce a novel approach that leverages sequential Bayesian updating to help students better allocate their attention to a subset of schools at which they have a higher probability of admission. Using China's gaokao market as a case study, we compare the performance of this Bayesian recommendation mechanism against that of the Chinese Parallel Binning mechanism as deployed in Zhejiang in 2017. Without sacrificing outcome stability, we find that our Bayesian approach generates significant welfare gains, across all levels of preference correlation, by drastically reducing the number of schools to which students must apply.