The “Wisdom of Crowds” and predicting box office revenues
Predicting Box Office with and without Markets: Do Internet Users Know Anything? (https://www.sciencedirect.com/science/article/pii/S0167624513000206?via%3Dihub)
In class, we discussed the wisdom of crowds and how state prices are essentially weighted averages of the aggregate betting community’s beliefs. HSX, or the Hollywood Stock Exchange, is a marketplace of enthusiasts who buy and sell shares of movies in accordance with their beliefs on how a movie is going to perform in the box office.
The journal article discusses HSX’s accuracy, comparing the traders’ predictions of revenue to the actual box office results. Their key findings indicate that HSX and Derby (a similar game run through the Box Office Mojo website) operate prediction models that exceed the accuracy of explanatory film covariates (like budget, opening week theatres, starpower, genre, etc.). The Absolute Percentage Error of HSX’s predictions in the study was .368, which, relative to other prediction models, is quite impressive. In a marketing report, HSX cited that they had 83.6% accuracy in predicting revenues for the top ten films each week (which corresponds to more than 80% of weekly box office revenues).
Prediction markets are ubiquitous, but the implications of the study are interesting- the “state prices” in this scenario are the revenues, but technically, each of the bettors has a wealth of 0. No real money is spent on HSX’s site by participants, and yet, their independent beliefs on the success/failure of a movie lead to box office predictions that are relatively accurate. It would be interesting to see the accuracy of the members on the leaderboards- a significant portion of the week-to-date leader board was populated with accounts active since prior to 2010. They have no financial incentives to participate on the site, so the relevance (or lack) of potential financial reward has an interesting implication in this context.