Matthew Rabin (Harvard) will present:
"Excess Movement in Belief Streams"
at 3pm on Wednesday, October 8, 2014 in 310 Silsby

You may view his schedule at the link below:
https://docs.google.com/spreadsheet/ccc?key=0As0WQNxSJWZVdDRPOGtCTGdwSkdxX2hpSm1vN3dGU0E&usp=sharing


                                                                                    Abstract
In this paper, we develop a new method to test if changes in beliefs are consistent with Bayesian updating based on correct priors. Our tests take advantage of a universal property of any Bayesian process that derives from its status as a bounded martingale: in expectation, the sum of squared changes of beliefs over a resolving belief stream, must equal the initial level of uncertainty. For example, for any data generating process with a prior of .4, the expected sum of squared changes across all potential signal realizations must equal .24. This principle allows for identi...cation of departures from Bayesian reasoning in some settings- observing whether people are "over-updating" or "under-updating,"- even without knowing what information people are observing or its normative meaning. For large data sets of observed belief streams, we provide both theoretical arguments and numerical illustrations that, despite its agnosticism about information structure, this test can often prove the presence of non-Bayesianness better than other potential tests, such as an unconditional martingale test, an autocorrelation test, and a variance-ratio test. We show how some known psychological biases  are likely to lead to too much or too little movement, and show some empirical patterns from a large set of sports-betting markets.




########################################################################

To unsubscribe from the ECONOMICS-SEMINARS list, click the following link:
https://listserv.dartmouth.edu/scripts/wa.exe?SUBED1=ECONOMICS-SEMINARS