Modeling for Broadway Success

Get ready data nerds – the data team at The Economist created an awesome interactive graph that illustrates a statistical model prediciting a Broadway production’s revenue (either weekly or cumulative over the first 9 months of performances).

Some of the variables you get to manipulate include the type of show (play or musical), the edition (original or revival), the genre, size of cast, size of theater, quality of reviews, and which lead actor you would like.

I have a certain fondness for this kind of ambitious analysis, as I did my senior thesis on this very topic (“Moneyball on Broadway:  A Statistical and Economic Exploration of How to Succeed on Broadway without Really Trying”). The challenges I faced doing this project were the data collection and cleaning, the new metrics one had to create in order to measure seemingly unmeasurable information (i.e. a celebrity’s popularity or value), and the regressions one has to run and rerun to ensure the correlations are significant.

The Economist’s model is built on a strong methodology, one which produced some surprising findings we as an industry should consider carefully.

  1. The New York Times review is not as impactful as a USA Today review.
    This finding was primarily from a paper written by NYU statistican Jeffery Simonoff, in which he assigned favorability scores from one to five for hundreds of Broadway reviews.  Mr. Simonoff and his colleagues “found that while shows that received positive reviews in USA Today and the Daily News tended to survive longer on Broadway, reviews in the New York Times has no statistically-significant effect”.
  1. How should you measure a Star’s value?
    This analysis used The Ulmer Scale. This is an index which measures “the perception of actors′ ‘bankability’ by assessing how easy it is to secure financing for a film starring them”. I think this is a tough measurement tool, as it assumes movie investors have the keen insight on consumer demand for a particular celebrity. There are better tools out there that measure news mentions, movie ratings, and social sentiment which indicates a celebrity’s popularity in a more linear way than assuming movie investors know best. A couple of tools to consider are iMDB STARmeter and the Hollywood Stock Exchange.
  1. Cast size is a real variable to consider
    I found a similar correlation in cast size in my analysis, that the larger the cast is, the higher the weekly gross should be. Why? One reason could be that the show is perceived more valuable with a larger cast. With ticket prices eclipsing $150 in some cases, the larger casts for the same price could seem to be better overall value on your dollar or it could seem to produce a better chance of having a fun time, justifing the high cost.

While this is overall a great model and fun chart, I have my few qualms with this analysis as well. I believe the theater size variable should be controlled for, as not all producers get the luxury of choosing their theater capacity. There is no mention of adjusting the grosses for inflation over time or utilizing economic measurements to control for US economy swings over time, which is an oversight. There is no definition on “major Tony awards” and there are other studies demonstrating how only Best Musical and Best Play are impactful awards on the bottom line.

I could go on, but you get the point – this kind of analysis is difficult and complex. 80% of Broadway shows do not recoup their investment and we sadly do not have a machine that can automatically create a profitable show at the push of a button (as fun as that may sound to my fellow data nerds and as cringeworthy of an idea as this may sound to my artistic colleagues). We may get there one day if we see more investment from the industry in conducting this kind of research in the right way and not doing it piecemeal using solely public data.

For now though, enjoy the chart. My 6-time Tony-winning original musical comedy (based on a Disney movie) starring Al Pacino in a cast of 90 people grosses $2.95M a week in a 1,938 seat house…all I need now is the capital.