Information Cascades and Movie Ratings
Positive online user-generated ratings can significantly increase the “buzz” surrounding newly launched movies, books, and other products. However, this buzz is incredibly fragile. Unrepresentative ratings of early customers can have long-term consequences on future customer purchase behavior. This is due to information cascades.
As we learned in class, an information cascade is a phenomenon in which individuals conform to others’ beliefs or actions even if their own judgment tells them otherwise. This type of observational learning lowers the quality of ratings created by users because their ratings tend to be strongly biased towards previous users’ ratings. Such an information cascade is problematic, leading to ratings that only reflect the perspectives of an initial set of customers. Thus, users who are consulting these ratings may not receive the full picture of each product because any conflicting perspectives on a product have been lost in favor of herding behavior.
Lee et al (2015) used data from a social networking movie community to investigate the relationship between user-generated movie ratings and observational learning. Based on model analysis, the researchers found that positive herding in user movie ratings tend to increase box office sales, while negative herding tend to hurt box office sales. However, the study revealed that social networking actually reduces the likelihood of herding on prior ratings. Because text reviews and other transparent forms of communications are more prevalent on social networking sites, the underlying reasons of people’s ratings are more visible to other users. Thus, herding behavior is moderated due to sharing of each user’s underlying reasons. This is in agreement with what we discussed in class. We are truly wiser as a crowd if we combine all of our unbiased observations. But if we do not share all of our observations, an information cascade may ensue.
https://repository.upenn.edu/cgi/viewcontent.cgi?article=1324&context=marketing_papers