Higher-order learning

Abstract We design a novel experiment to study how subjects update their beliefs about the of others. Three players receive sequential signals an unknown state world. Player 1 reports her state; 2 simultaneously 1; 3 2. say that exhibit higher-order learning if k $$k-1$$ k - 1 become more accurate as are observed. find some predicted dynamics reflected in data; particular, updated slowly with private than public information. However, fails even after large number is argue this result driven by base-rate neglect, heterogeneity updating processes, and subjects’ failure correctly take rules others into account.