Identification of the Determinants of the Films Box Office (Case Study of the USA Film Industry)

The film industry has generated considerable interest in terms of financial aspects since it has transformed into a business and seeks to generate profit. The box office as one of the revenue sources is considered to be a transparent and relevant indicator of the film market’s condition, particularly, in the United States – the largest film industry. The topic of forecasting box office receipts is quite relevant due to the lack of research and its relevant youth. A substantial pool of research is devoted to examining the determinants of box office receipts and predicting the box office. However, these approaches seem to be diverse and may not be comprehensive; therefore, it is necessary to supplement previous research on the topic of determining the factors of box office profitability of a film project. The study aims to determine the factors that affect the box office revenue and to predict the box office based on several independent variables. To determine such factors, the work suggests constructing a linear regression model and implementing models based on gradient boosting (XGBoost, CatBoost, and LightGBM). As one of the supposed determinants, the study considers the success rate of actors and directors, calculated by the author's method based on the box office of past film projects in which the actor or director took part. The results of the regression analysis show high significance of such variables as the production budget, the power of actors and directors, whether the film is a sequel or not, and several film genres. Gradient boosting models show significant predictive accuracy, also considering the above-mentioned variables as significant ones. Thus, the models proposed in the study can have both theoretical application in the methodological point of view, and practical application for film producers who aim at predicting the profitability of their project. Keywords: box office, film industry, forecasting, gradient boosting, linear regression