Handbook of Media Economics, vol 1B

of Yelp ratings, which endogenously provides more weight to later reviews, and to elite

reviewers, among other factors.

At the opposite extreme of arithmetic averages is the approach of finding “bad” con-

tent and removing it all together. Platforms that use algorithms to identify and remove

content thought to be fake use this approach, as do spam detection algorithms (for exam-

ple, in

Ott et al., 2011). As with the case of arithmetic averaging, the approach of remov-

ing content altogether is only optimal under very restrictive assumptions, as it assumes

that the removed content contains no useful information. Taking a step back from this

specific application, it seems that, given some model of reviewer behavior, it should be

possible to derive