rankings

Bayesian Social Learning, Conformity, and Stubbornness: Evidence from the AP Top 25 : 

Bayesian Social Learning, Conformity, and Stubbornness: Evidence from the AP Top 25 Discussion

The plan : 

The plan Objective: improve estimate of college-football ranking by as much as possible. Proxy for best estimate: voter’s own season-ending rankings. Information production: each week’s results and aggregate rankings lead to updated rankings by each voter. Research question: to what extent are updates rational?

Slide 3: 

College football presents an unusually poor data set for measuring learning over the season, because of small number of games, the way ranked teams strategically avoid scheduling strong nonconference opponents, and the fact that there is no playoff system. In addition, it is not clear that football rankings are obviously transitive. The assumption that final scores are the only in-game information needed to update rankings, if false, means that estimates of Bayesian learning are biased downward.

Slide 4: 

If voters care about their reputation, why isn’t matching the season-ending aggregate rankings the objective?

Slide 5: 

Another method exists to test financial effects of reputation, namely polls in less lucrative sports. These sports, having playoffs, also provide perhaps a better measure of the accuracy objective than college football. In particular, predictions of NCAA basketball tournament outcomes at Yahoo, etc. often allow voters to update their brackets after each round. Here, the outcome to be best estimated – the final tournament results in all rounds – is obvious and uncontroversial.

Slide 6: 

Why use AggB and AggW, which are crude measures of social information? Why not use difference between voter’s and aggregate ranking?

Slide 7: 

The fact that inexperienced voters respond more strongly to social information could be a rational acknowledgment of poor information as much as a greater concern for reputation.

Slide 8: 

Are gaps between actual and Bayesian responses to social rankings smaller later in season?

Slide 9: 

Do nationally televised games generate more under-response to social information? If voters are stubborn, I expect they would.

Minor stuff : 

Minor stuff Voter “tastes regarding true ranings” seems a peculiar phrase, and I’m not sure what it means. It’s not clear how “YTD performance” is different from “the best estimate of the rankings at the end of the season, based on what we know now.”