In December 2014, the Centre National du Cinéma (CNC) – the regulation body that in particular distribute subsidies – decided to enforce a ceiling to the revenues of actors participating in a movie, it that particular movie benefits from subsidies. This decision came in echo to a debate launched by Vincent Maraval in December 2012 in Le Monde: because of the “miracle of the French financing system for movies”, actors would be too much paid and thus budgets for French movies too high.
The heated debate that ensued between Vicent Maraval and his opponents revolved very much around a few examples chosen to impress the general public : Dany Boon, Gérard Depardieu, Astérix, Populaire,….To the dismay of statisticians, and as if genericity would be built from particularities.
It was thus quit obvious for us to try and see if a quantitative analysis could help shed some light on the debate. As in many other domains, lots of information is available on the web. You just need to spend a little (quite a lot, actually…) time to collect it. We collected data on 10 000 movies, viewed in French movie theatres since 2001: number of tickets sold, number of screens allocated to each movie by distributors, director, main actors, when it was first shown, budget, country, type, reviews in the press, ratings given by movie goers on Allo Ciné.
Our econometric model explains the total number of ticket sold as a function of the data just mentioned. The month when the movie is out is taken into account, as well as the fact the movie was possibly shown during the Olympics, the soccer Euro or World Cup. We also account for the average temperature in Paris, Lyon and Marseille, as well as data on the competitive environment: how many movies were out, or shown, at the same time.
A crucial variable in the model is the number of screens allocated by distributors. Whatever the quality of a movie, if it is shown only in one place, it will not sell many tickets… This number of screens incorporates the distributors’ perception of a movie, formed in particular from the same information that enter into the model, like its budget, its timing and of course its casting.
The impact on tickets sold as measured by the model is thus the part of the impact that is not anticipated by distributors. We include directors or actors in the model only if our sample contains at least 5 movies with them. If the model concludes to a significant impact, it means that distributors underestimate or overestimate, with some persistence, the impact of an actor (or a director) on the success of a movie,
Moreover, as we use a hierarchical Bayesian modelling, we can also assess the variability of an actor’s impact on a movie’s success. In other words, if the only notable feat of arms of an actor is one film that over performed (north, south, east and west), when all his others’ are below expectations, our model can spot that.
First learning of the model: distributors anticipate rather well the impact of actors on a film’s success. We have 388 actors in our model. Only 76 have a significant impact. Only for those 76 are distributors recurrently mistaken.
Second learning: distributors more often err on the undervaluation side than on the overvaluation one. Only 26 actors have a negative impact, which means that distributors overestimate sales when they are cast. Among them, on the French side, Christian Clavier, Gérard Jugnot or Audrey Tautou. In good company actually, as George Clooney or Leonardo DiCaprio also fall in that category.
Big surprise then when we look at actors undervalued by distributors. 18 in the top 20 are French: Sandrine Bonnaire, Nicole Garcia, Yolande Moreau, Louis Garrel, Michel Serrault, Charles Berling, Fanny Ardant, Esla Zylberstein ou Isabelle Huppert are among them. Thus, in our analysis, nothing seems to confirm that French actors are cost too much for their worth.
It is also interesting to look at other impacts. Concerning press reviews, Positif, Télérama, Télé 7 jours and La Croix all have a significant impact on tickets sold. When it comes to ratings posted on Allo Ciné, average ratings are not those that matter. Low ratings given in the first two weeks of the life of a movie do have an impact, as well as high ratings all along. And finally, in a crowded and competitive market, over supply has a negative impact on tickets sold that distributors seem not to be able to anticipate.