Green-lighting Movie Scripts: Revenue Forecasting and Risk Management

Major studios and independent production firms (Indies) often have to select or “greenlight” a portfolio of scripts to turn into movies. Despite the huge financial risk at stake, there is currently no risk management tool they can use to aid their decisions, even though such a tool is sorely needed. In this paper, we developed a forecasting and risk management tool, based on movies scripts, to aid movie studios and production firms in their green-lighting decisions. The methodology developed can also assist outside investors if they have access to the scripts. Building upon and extending the previous literature, we extracted three levels of textual information (genre/content, bag-of-words, and semantics) from movie scripts. We then incorporate these textual variables as predictors, together with the contemplated production budget, into a BART-QL (Bayesian Additive Regression Tree for Quasi-Linear) model to obtain the posterior predictive distributions, rather than point forecasts, of the box office revenues for the corresponding movies. We demonstrate how the predictive distributions of box office revenues can potentially be used to help movie producers intelligently select their movie production portfolios based on their risk preferences, and we describe an illustrative analysis performed for an independent production firm.