Why the Oberholzer-Gee/Strumpf Article on File Sharing Is Not Credible

This article examines the data, results, and methods underlying an influential 2007 article on music piracy published in the Journal of Political Economy. The article concluded that piracy had no impact on record sales, even though the birth of file sharing coincided with what in hindsight can be described as a financial near-collapse of the sound recording industry. The authors of that article had access to music download data from actual pirate servers—data that the authors have never made public—providing much of the article’s novelty and lending it an aura of authenticity. They also relied upon an esoteric instrument, the number of German kids on school vacations, which they claimed had a powerful effect on American piracy. My examination identifies several important concerns and problems. First, the measured aggregate piracy data contain extremely large weekly variations that are inconsistent with other, comparable data. Then, the first-stage instrumented regression results imply, counterfactually, that American piracy is dominated by whether German children are in school or not. Further, their reported values erroneously indicate that German students spend most of their time on holiday, indicating an error in the construction of that variable. In addition, the weekly aggregate relationship between American piracy and German school vacations is the opposite of the hypothesized relationship underlying the use of the instrument. Most importantly, a careful investigation reveals that the extra files available to Americans because of German school holidays made up less than two-tenths of one percent of all files available to Americans, which is a much smaller share than the article authors suggest and far too small to have had a measurable impact on American piracy.