## Monday, August 28, 2006

### Movie Reviews, stars, box office revenue and normal probability distribution

One point I failed to take up in the post below was just what can be used to predict box office revenue. Luckily my masters thesis focuses on the impact critical reviews have on movie revenues (my undergraduate paper focused on in theatres while my masters extended and improved on a data set for video rentals). Essentially what the data tells us is the following:

Trying to come up with a formula for producing movies will not work. A feature of box office reviews is that there is no normal probability distribution---which is a fancy term for saying that while there is a mean (say \$100 million a movie) there is an infinite variance. When a data set has an infinite variance that means that most of the data is all over the place—one movie at 1 million, another at 800 million—and you can’t pinpoint exactly where the data will fall. For instance, in Stats101 you assume there is a “finate” variance and then you calculate what is called a standard deviation, which can be used to tell you how much data falls within a certain range of the mean.

This complicates any statistical analysis , but I tried to run what is called OLS on data available free online. My tentative conclusions—which I updated using 2001-2002 data)—is as follows.

1) The NY Times article is right—star power has little effect on overall box office returns.
2) Other factors, such as being a franchise, help you open a movie for the initial weekend but will not increase your returns much afterwords.
3) Critical reviews—which in my dataset is a compilation of all online and newspaper reviews as a % positive—are technically significant but will not make or break a movie. Overall, for instance, in increase from 50% positive to 70% positive will give the movie studio an extra \$10mn
4) Breaking down the effect of critical reviews show that the play both a small effect in the opening weekend and a decent effect on the amount of time the movie remains in the theatre. Using the same example from above, the 20% better reviews adds an extra \$3.25mn to box office revenue while extending the amount of time in the theatre 11 days.
5) Keeping in line with the article and the relevant literature, a bad review is a better predictor than a good review. That means reviews can kill a movie but they will not make one on their own.
6) Given the incentives of the movie industry (note below that the production company takes a much bigger share of the opening week revenue total than they do subsequent weeks), shorter term factors are more profitable. For instance, a movie that is a franchisewill add \$13.6mn during the opening weekend.But despite all this work I did for my masters thesis, the essential fact remains, it is almost impossible to predict what movies will make. Sometimes being a franchise will help (Pirates of the Caribbean?), but sometimes it won’t (Speed 2 anyone?).Yet given the almost infinite variance of movie revenue, perhaps having a Tom Cruise, a popular adaptation or a franchise will help lower the risk in making the movie.

#### 1 comment:

Ayo Sekolah said...