MARCH MADNESS? UNDERREACTION TO HOT AND COLD HANDS IN NCAA BASKETBALL

摘要

The hot hand bias is the widely documented bias toward overestimation of positive serial correlation in sequential events. We test for the hot hand bias in a novel real-world context, the seeding of National Collegiate Athletic Association (NCAA) basketball tournament teams. That is, we examine whether teams that perform relatively well heading into March Madness are seeded too high, and/or teams with poor recent performance are seeded too low. The seeds are determined by a 10-member committee that only has implicit incentives, but these incentives are still substantial as the committee's decisions are highly scrutinized by the media, fans, and other stakeholders. We find that, contra the hot hand bias, the committee underreacts to signals of momentum heading into the NCAA tournament. Various results indicate this behavior can be attributed to both: (1) inattention to relatively detailed information indicating momentum; and (2) under-appreciation of the predictive value of this information. Betting markets incorporate this information efficiently, but neglect some additional information that is predictive of winning NCAA tournament games but not of beating the spread. We note that the NCAA tournament has been highly popular and lucrative partly due to the madness (high frequency of wins by lower-seeded teams), which the bias we document contributes to, making the persistence of bias less surprising. (JEL D83, L83)