A two-stage proxy variable approach to estimating movie box office receipts

Abstract

This paper advances the ongoing discussion of methods for predicting movie box office revenues with two contributions to the methodology and an out-of-sample test of the model. The first innovation is the development of a two-stage model using publicly available pre-release indicators to predict (1) initial week and (2) subsequent run box office revenues. To incorporate the experience-good nature of movies, the second stage is estimated by incorporating a proxy variable for box office success during the first week relative to predicted first week success. The second contribution is an empirical test of De Vany and Walls’ (J Econ Dyn Control 28:1035–1057, 2004) finding that the distribution of movie revenues has “heavy tails” and follows a non-Gaussian stable distribution with infinite variance. We estimate the two-stage model of a movie’s box office success on all general release movies in 1 year with both the Gaussian and stable distribution with heavy tails and infinite variance and find no evidence for the stable distribution in either stage of the estimation. This two-stage model is validated by comparing all general release movies in 3 future years (out-of-sample data) to the model’s predictions.