Empirical data on file-sharing's effect on album sales

Koleman Strumpf, a conservative, Cato-affiliated economist at the University of North Carolina at Chapel Hill, has just co-authored a paper on the effects of file-sharing on album sales, based on the first-ever empirical data analysis in the field. Koleman watched the file requests on OpenNap servers (to get numbers on which albums' tracks are being downloaded) and compared them to the sales-figures for each album, correlating file-sharing popularity against sales data. His conclusion: file-sharing isn't killing record sales.
We analyze a large file sharing dataset which includes 0.01% of the world’s downloads from the last third of 2002. We focus on users located in the U.S. Their audio downloads are matched to the album they were released on, for which we have concurrent U.S. weekly sales data. This allows us to consider the relationship between downloads and sales. To establish causality, we instrument for downloads using technical features related to file sharing (such as network congestion or song length) and international school holidays, both of which are plausibly exogenous to sales. We are able to obtain relatively precise estimates because the data contain over ten thousand album-weeks...

Even in the most pessimistic specification, five thousand downloads are needed to displace a single album sale...high selling albums actually benefit from file sharing.

369K PDF Link (Thanks, Koleman!)