Abstract:
This paper empirically analyzes the effect of the widely used agency model on the retail prices of e-books in United Kingdom. Using a unique cross-sectional data set of e-book prices for a large sample of book titles across all major publishing houses, we exploit cross-genre and cross-publisher variation to identify the mean effect of the agency model on e-book prices. Since the genre information is ambiguous and even missing for some titles in our original dataset, we use a Latent Dirichlet Allocation (LDA) approach to determine detailed book genres based on the book's descriptions. We find that e-book prices for titles that are sold under the agency model are 36% cheaper than titles sold under the wholesale model on average. Our results are robust to different specifications, a Lewbel instrumental variable approach, and machine learning techniques.