The Effects of Latent Categories on Product Performance: Discovery, Identification and Firm Strategy in the Context of the Motion Picture Industry

My dissertation, composed of three chapters, explores the effect of product categories, as revealed by consumers' online rental behavior, on firm performance and strategy in the movie industry. Research in sociology suggests that sorting products into cognitive categories matters for performance, but empirical studies find an inconsistent relationship between category-spanning and performance. Importantly, they do not establish the causality of this hypothesized relationship. To address these issues, the first chapter empirically identifies the relationship between category-spanning and movie performance, based on a measure of latent category coherence, entitled latent similarity, which is adapted from Zuckerman (2004). This represents a departure from existing research that relies on third-party categories to measure category-spanning, as this may not capture how consumers categorize. I evaluate a movie's latent similarity from consumers' movie rental patterns on Amazon Instant Video, which are observable in movies' "Customers who rented this also rented..." co-rental lists. The idea behind the measure is that if all of a movie's co-rentals cross-reference each other (forming closed set), then that movie has a high latent similarity measure, suggesting it is in a clear category. I find a strong positive relationship between the measure and performance, and I provide causal evidence through an identification strategy composed of instrumental variable and control function. The second chapter explores what exactly the latent similarity measure is picking up. I identify the latent categories implied by the Amazon rental data through clustering algorithms. Then, I employ topic modeling to show that the derived categories differ in the extent to which their movies share common attributes, and I develop a theoretical model to explain how consumer preferences can lead to these differences. The third chapter is an empirical investigation of studio strategies regarding their movies' category-spanning, exploring the role played by studio type in this decision. My dissertation has several contributions. First, it provides causal evidence that variation in category coherence affects performance. Second, it provides several methodological innovations for strategy in (1) adapting the Zuckerman measure to use with novel digital data, and (2) applying machine learning tools to identify and describe latent product categories.Ph.D