Writing for Cato At Liberty, Ars Technica alum Julian Sanchez has a timely redux of the research he did on how the made-up piracy numbers quoted during debates about SOPA and PIPA come from, and how little relation they bear to reality. It seems like every discussion of SOPA/PIPA includes a phrase like "Everyone agrees that piracy is huge problem," but in fact, the "huge problem" they're agreeing on has been inflated to farcical proportions through the most transparent financial funny business.
Siwek takes an estimate of $6.1 billion in piracy losses to the U.S. movie industry, and through the magic of multipliers gets us to a more impressive sounding $20.5 billion. That original $6.1 billion figure, by the way, was produced by a study commissioned from LEK Consulting by the Motion Picture Association of America. Since even the GAO was unable to get at the underlying research or evaluate its methodology, it's impossible to know how reliable that figure is, but given that MPAA has already had to admit significant errors in the numbers LEK generated, I'd take it with a grain of salt.
Believe it or not, though, it's actually even worse than that. SOPA, recall, does not actually shut down foreign sites. It only requires (ineffective) blocking of foreign "rogue sites" for U.S. Internet users. It doesn't do anything to prevent users in (say) China from downloading illicit content on a Chinese site. If we're interested in the magnitude of the piracy harm that SOPA is aimed at addressing, then, the only relevant number is the loss attributable specifically to Internet piracy by U.S. users.
Again, we don't have the full LEK study, but one of Siwek's early papers does conveniently reproduce some of LEK's PowerPoint slides, which attempt to break the data down a bit. Of the total $6.1 billion in annual losses LEK estimated to MPAA studios, the amount attributable to online piracy by users in the United States was $446 million—which, by coincidence, is roughly the amount grossed globally by Alvin and the Chipmunks: The Squeakquel.