By Kaylyn Kluck
Hollywood investors may need to start paying more attention to University of Iowa research.
Since the summer of 2013, UI Assistant Professor of business Kang Zhao and UI graduate student Michael Lash have tried to find a reliable way to predict the profitability of film releases. Their shared interests in movies and big data inspired them to develop a unique way for determining box-office success.
Nothing like this has been studied before; no one has tried to predict the profitability of movies, and most studies try to predict the success of a film very late in production, when there is little an investor can do, Zhao said.
“What we’re trying to do, from a research perspective, is predict the profitability of movies at a very early stage,” he said.
The patterns in the data revealed the biggest factor in movie profitability is pairing actors and directors who have a history of making successful films together.
“The previous profitability record of actors certainly matters, but it’s not as important as the team perspective,” Zhao said.
The second most important factor is the director’s ability to bring revenue to a movie.
“So you say JJ Abrams or you say Steven Spielberg, and I think that has a large effect on the crowd that’s going to see a movie,” Lash said.
Zhao and Lash define a profitable movie as one that makes more than $7.3 million in profit. They found that segmenting a movie’s audience, such as through age restrictions, is the biggest way to hurt profitability. Drama, war, and foreign films are less likely to turn a profit, and rating is also a factor.
“An R-rating is limiting in the sense that not everyone can go see it, and maybe not everybody wants to go see that,” Lash said.
The researchers have predicted what movies will be successful in 2016. Their top pick is the upcoming Disney release Zootopia, which the algorithm says has a greater than 50 percent chance of being profitable.
“We looked at the ‘what’s hot’ each year sort of factor,” Lash said. “The animation genre has been hot lately with Frozen and stuff like that.”
Nick Street, the head of the Tippie College of Business Management Sciences Department, said Zhao and Lash’s study represents what their researchers are able to do.
“This is a nice example of data-mining computer science,” he said, “People working on specific business-related problems in a way other people don’t know how to predict.”
Because manually copying and pasting data from thousands of movies would be impossible, Lash had to create a code that would automatically collect statistics.
“I built what’s called a web scraper to scrape lots of data from a site called Box Office Mojo, and later augmented that to scrape additional information from IMDB,” he said. “We now have, I think, morethan 16,000 or 17,000 movies in our data base.”
The researchers said they used an algorithm called “random forest” to predict profitability.
“We outsourced this prediction job to tens of thousands of prediction trees, and had them decide what is the best outcome,” Zhao said.
So far, they’ve found their algorithm to be pretty accurate with movies that have been released in 2016.
“I looked at the freely-available information that the internet had circulated on movies that had already come out, and I was like, OK, it looks like we’re doing pretty good,” Lash said.
The researchers said a few firms in Hollywood have already taken an interest in their findings.
“It does have potential to be used for purposes other than research,” Zhao said. “We’re just trying to do it for fun. We have more information than we could use.”