When the Institute for Policy Innovation published a study purporting to show the harms of movie piracy to the United States economy, I wrote a post critiquing it that was unnecessarily vituperative. After further reflection, I posted a follow-up post apologizing for the tone of that post. IPI president Tom Giovanetti apparently didn’t find my apology adequate, because he sent letters to the president and several board members of the Show-Me Institute, where I was employed at the time, seeking to have me reprimanded. Thankfully, they didn’t consider my post to be a firing offense.
I’d like to avoid repeating that experience, so I’m going to be a lot more polite in my analysis of IPI’s latest study, this one on music piracy. Unfortunately, the new paper exhibits the same methodological defects as the previous study, and introduces some new problems as well. The gory details are below the fold.
The fundamental problem is the same one made by the first study: it not only counts the direct revenues of the music industry, but also the “cascade” of additional revenues that flow to the labels’ suppliers, such as recording studios, jewel case manufacturers, etc.
This is double counting. If a foreigner gives me $1, and I turn around and buy a plum from you for a dollar, and then you turn around and buy a peach from another friend for a dollar, we haven’t thereby increased our national wealth by $3. At the beginning of the sequence, we have a plum and a peach. At the end, we have a plum, a peach, and a dollar. Difference: one dollar. No matter how many times that dollar changes hands, there’s still only one dollar that wasn’t there before.
Yet using the methodology of Siwek’s study, when a record label makes $10 selling a CD to a Canadian, and then gives $7 to the company that manufactured the CD and $2 to the guy who shipped it to Canada, society has benefitted by $10+$7+$2=$19. Yet some simple math shows that this is incorrect: the studio is $1 richer, the trucker is $2, and the manufacturer is $7. Shockingly enough, that adds up to $10. What each participant cares about is his profits, not his revenues.
And why stop there? The trucker would have had to fill up his truck with fuel in order to drive the CD to Canada. Why not count that transaction as a cost of piracy? And in order for the gas station to sell the trucker the fuel, it would have had to purchase it from a refinery. That refinery’s lost gasoline revenues are also a cost to piracy. And what about the gas pump? The gas station had to buy it from somebody! A fraction of the gas pump manufacturer’s revenues are attributable to that CD sale. And of course, to manufacture the gas pump, they had to buy parts from a bunch of other companies. They too… well, you get the idea.
I won’t bore you with the math as I did with the movie study, but suffice it to say that of the $12.5 billion in claimed lost output, $7.2 billion are double-or-triple-counted revenues from industries providing inputs to the recording industry, while only $5.3 billion is actual losses to the recording industry. Only the latter represents a net reduction in Americans’ wealth. The other $7.2 billion is “lost output” only in the sense that unbroken windows represent “lost output” for glaziers.
And even the $5.3 billion figure strikes me as inflated. They assume that 20 percent of peer-to-peer downloaded music would otherwise have been sales—an optimistic estimate, I think, but not a crazy one. But then they assume that the songs would have sold for an average of $2.31 apiece. That’s because he assumes that 90% of purchasers would have paid 99 cents for a download, and the other 10 percent would have paid an average of $14 to buy the CD containing the song.
That doesn’t make a lot of sense to me. Presumably, people don’t usually purchase an entire CD just to get one song. Conversely, one assumes that in many cases the users of peer-to-peer applications download more than one song from a given album. In either case, assuming that the average sale price of a song is $2.31 seems to me to overstate things by at least a factor of 2.
If we adjust the per-song revenue estimate to a more plausible figure of 99 cents, the total direct losses to the music industry drops from $5.3 billion to $3.2 billion.
In total, then, two questionable assumptions increased their estimate by nearly a factor of 4, from $3.2 billion to $12.5 billion. And even that figure may be an over-estimate because this study relies on survey data collected by an industry trade group. And as we’ve written before piracy surveys conducted by industry trade groups often have serious methodological problems and should be taken with a grain of salt.
The Technology Liberation Front is the tech policy blog dedicated to keeping politicians' hands off the 'net and everything else related to technology.