By Julian Sanchez Published: October 07, 2008 - 11:30PM CT from Ars Technica
A 20-year game of Telephone
If you pay any attention to the endless debates over intellectual property policy in the United States, you'll hear two numbers invoked over and over again, like the stuttering chorus of some Philip Glass opera: 750,000 and $200 to $250 billion. The first is the number of U.S. jobs supposedly lost to intellectual property theft; the second is the annual dollar cost of IP infringement to the U.S. economy. These statistics are brandished like a talisman each time Congress is asked to step up enforcement to protect the ever-beleaguered U.S. content industry. And both, as far as an extended investigation by Ars Technica has been able to determine, are utterly bogus.
"I have said it thrice," wrote Lewis Carroll in his poem The Hunting of the Snark, "what I tell you three times is true." And by that standard, the Pythagorean Theorem is but schoolyard gossip compared with our hoary figures. As our colleagues at Wired noted earlier this week, the 750,000 jobs figure can be found cited by the U.S. Department of Commerce, Customs and Border Patrol, and the U.S. Chamber of Commerce, among others. Both feature prominently on TheTrueCosts.org, an industry site devoted to trumpeting the harms of piracy. They're invoked by the deputy director of the U.S. Patent and Trademark Office. And, of course, they're a staple of indignant press releases from the congressional sponsors of tough-on-piracy legislation.
By more conventional standards of empirical verification, however, the numbers fare less well. Try to follow the thread of citations to their source, and you encounter a fractal tangle of recursive reference that resembles nothing so much as the children's game known, in less-PC times, as "Chinese whispers," and these days more often called "Telephone." Usually, the most respectable-sounding authority to cite for the numbers (the FBI for the dollar amount, Customs for the jobs figure) is also the most prevalent—but in each case, that authoritative "source" proves to be a mere waystation on a long and tortuous journey. So what is the secret origin of these ubiquitous statistics? What doomed planet's desperate alien statisticians rocketed them to Kansas? Ars did its best to find the fountainhead. Here's what we discovered.
Looking for lost jobs
First, the estimate of 750,000 jobs lost. (Is that supposed to be per year? A cumulative total over some undefined span? Those who cite the figure seldom say.) Customs is most often given as the source for this, and indeed, you can find press releases from as recently as 2002 giving that figure as a U.S. Customs and Border Patrol estimate. Eureka! But when we contacted CBP to determine how they had arrived at that imposing figure, we were informed that it was, in essence, a goof. The figure, Customs assured us, came from somewhere else, and was mistakenly described as the agency's own. This should come as no great surprise: CBP is an enforcement agency, whereas calculating the total loss of jobs from IP infringement would require some terrifyingly complex counterfactual modeling by trained economists. Similar claims have appeared in Customs releases dating back at least to 1993, but a CBP spokesperson assured us that the agency has never been in the business of developing such estimates in-house.
With Customs a dead end, we dove into press archives, hoping to find the earliest public mention of the elusive 750,000 jobs number. And we found it in—this is not a typo—1986. Yes, back in the days when "Papa Don't Preach" and "You Give Love a Bad Name" topped the charts, The Christian Science Monitor quoted then-Commerce Secretary Malcom Baldridge, trumpeting Ronald Reagan's own precursor to the recently passed PRO-IP bill. Baldridge estimated the number of jobs lost to the counterfeiting of U.S. goods at "anywhere from 130,000 to 750,000."
Where did that preposterously broad range come from? As with the number of licks needed to denude a Tootsie Pop, the world may never know. Ars submitted a Freedom of Information Act request to the Department of Commerce this summer, hoping to uncover the basis of Baldridge's claim—or any other Commerce Department estimates of job losses to piracy—but came up empty. So whatever marvelous proof the late secretary discovered was not to be found in the margins of any document in the government's vaults. But no matter: By 1987, that Brobdignagian statistical span had been reduced, as far as the press were concerned, to "as many as 750,000" jobs. Subsequent reportage dropped the qualifier. The 750,000 figure was still being bandied about this summer in support of the aforementioned PRO-IP bill.
$250 billion? What's that in real money?
What, then, of that $200 to $250 billion range? Often, it's attributed to the Federal Bureau of Investigation, and indeed, the Bureau routinely cites those numbers. According to FBI spokesperson Catherine Milhoan, the figure "was derived through our coordination with industry, trade associations, rights holders, and other law enforcement agencies" at a 2002 anti-piracy confab. But neither the Bureau nor the National Intellectual Property Rights Coordination Center, which assembled the inter-agency powwow, could find any record of how that number was computed.
At this point, it's necessary to get a little speculative. As with Customs, the FBI is not in the habit of doing sophisticated economic analysis in-house. And the last time the government conducted any sort of verifiably rigorous study of the costs of IP theft—about which more presently—it was a protracted undertaking that involved sending detailed questionnaires to hundreds of businesses, which government economists concluded was still insufficient to produce a reliable figure for the economy as a whole. However, $250 billion is about the number you come up with if you start with $200 billion in 1993 dollars and adjust for inflation to 2002. And that lower end of the range, $200 billion, happens to date back to 1993.
Another group that routinely uses the $200 to $250 billion figure is the International Anti-Counterfeiting Coalition, which (along with the FBI) is often given as the source of the number. That organization's white papers, as recently as 2005, footnote the figure to 1995 congressional testimony urging passage of what became the Anticounterfeiting Consumer Protection Act of 1996. So Ars dug into the archives at the Library of Congress to discover where the witnesses before the House and Senate Judiciary Committees got their data.
Several of the witnesses were conspicuously vague about their sources. An IACC factsheet submitted for the hearings said the group itself "estimates the economic cost due to product counterfeiting to exceed $200 billion each year," a number repeated by the group's then-president, John Bliss. Congressman Bob Goodlatte (R-VA) gave the same figure without sourcing. But several witnesses pointed to Forbes magazine as the source of the number. Rep. John Conyers (D-MI) noted that the International Trade Commission had placed the size of the counterfeit market at $60 billion in 1988 and that "a more recent estimate by Forbes Magazine says that American businesses are losing over $200 billion each year as a result of illegal counterfeiting." Finally, Charlotte Simmons-Gill of the International Trademark Association was kind enough to give a precise citation: the October 25, 1993 issue of Forbes.
Ars eagerly hunted down that issue and found a short article on counterfeiting, in which the reader is informed that "counterfeit merchandise" is "a $200 billion enterprise worldwide and growing faster than many of the industries it's preying on." No further source is given.
Quite possibly, the authors of the article called up an industry group like the IACC and got a ballpark guess. At any rate, there is nothing to indicate that Forbes itself had produced the estimate, Mr. Conyers' assertion notwithstanding. What is very clear, however, is that even assuming the figure is accurate, it is not an estimate of the cost to the U.S. economy of IP piracy. It's an estimate of the size of the entire global market in counterfeit goods. Despite the efforts of several witnesses to equate them, it is plainly not on par with the earlier calculation by the ITC that many had also cited.
But here, at last, we have a solid number to sink our claws into, right? Sure, it's 20 years old, but the U.S. International Trade Commission at least produced a reputable study yielding a definite figure for the cost of piracy to the U.S. economy: $60 billion annually.
Well, not quite.
"Biased & self-serving"
The number the ITC actually came up with, based on a survey of several hundred business selected for their likely reliance on IP for revenue, was $23.8 billion—the estimated losses to their respondents. That number was based on industry estimates that the authors of the study noted "could admittedly be biased and self-serving," since the firms had every incentive to paint the situation in the most dire terms as a means of spurring government action. But the figures at least appeared to be consistent and reasonable, both internally and across sectors.
The $60 billion number comes from a two-page appendix, in which the authors note that it's impossible to extrapolate from a self-selecting group of IP-heavy respondents to the economy as a whole. But taking a wild stab and assuming that firms outside their sample experienced losses totaling a quarter to half those of their respondents, the ITC guessed that the aggregate losses to the economy might be on the order of "$43 billion to $61 billion."
The survey also, incidentally, asked respondents to estimate the number of job losses they could attribute to inadequate intellectual property protection. The number they came up with was 5,374. If we assume, very crudely, that job losses are proportionate to dollar losses, then the ITC's high-end estimate of $61 billion in total economic costs would correspond to a loss of not 750,000 jobs, but 13,774.
If we want to be very precise, however, we should note that the ITC was not calculating losses from IP "theft," but rather "inadequate protection" of intellectual property. And "inadequate protection" was interpreted to mean protection falling short of the level provided by U.S. law. The protection provided by a foreign country might be deemed "inadequate," the study explained, if "exceptions to exclusive rights are overly broad"—for example, if a country's law contained "broad exceptions for public performances in hotels or film clips" or "too broad exceptions for educational photocopying." A legal regime could be "inadequate" because "terms of protection are too short" or because of "inadequate" civil or criminal remedies, meaning monetary damages or criminal penalties for infringers were not high enough.
Calculating the net cost of piracy to the economy
One final, slightly theoretical point deserves emphasis here. All the projections we've discussed, the rigorous and the suspect alike, calculate losses in sales or royalties to U.S. firms. This is often conflated with the net "cost to the U.S. economy." But those numbers—whatever they might be—are almost certainly not the same. When someone torrents a $12 album that they would have otherwise purchased, the record industry loses $12, to be sure. But that doesn't mean that $12 has magically vanished from the economy. On the contrary: someone has gotten the value of the album and still has $12 to spend somewhere else.
In economic jargon, charging anything for pure IP—which has a marginal cost approaching zero once it has been produced—creates a deadweight economic loss, at least in static terms. The actual net loss of IP infringement is an allocative loss that only appears in a dynamic analysis. Simply put, when people pirate IP, the market is not accurately signaling how highly people value the effort that was put into creating it, which leads to underproduction of new IP. To calculate the net loss to the economy over the long run, you'd need to figure out the value of the lost innovation in which IP owners would have invested the marginal dollar lost to piracy, and subtract from that the value of the second-best allocation—which is to say, whatever the consumer of the pirated good spent his money on instead—and the value of the deadweight loss (free music or software is a net economic benefit to someone) incurred by pricing IP at all.
If that sounds incredibly complicated, it is. And in fact, it's more complicated than that, because as Yochai Benkler has argued persuasively, IP is an input to innovation as well as the product of innovation. So under certain very specfic conditions, "piracy" can produce net gains. While it seems extremely unlikely that this is the case in the aggregate—IP theft almost certainly does impose net economic costs—simply calculating lost sales and licencing fees, assuming someone could produce a credible figure, would not provide a complete picture of the economic impact of IP infringement. It would give us, at most, one side of the ledger.
Conclusions
But enough theory and speculation; here is what we can say for certain: the two numbers that are invariably invoked whenever Congress considers the need for more stringent IP enforcement are, at best, highly dubious. They are phantoms. We have no good reason to think that either is remotely reliable.
Perhaps more importantly, both numbers are seemingly decades old, gaining a patina of currency and credibility by virtue of having been laundered through a relay race of respectable sources, even as their origin recedes into the mists. That's especially significant, because these numbers are always invoked as proof that the piracy problem is still dire—that everything we've done to step up international enforcement of intellectual property laws has been in vain. But of course, if you simply recycle the same numbers from 15 and 20 years ago—remember that IACC's 2005 publications still cite that 1995 congressional testimony, from which it seems safe to infer that they have no more recent source—then it will necessarily seem as though no ground has been gained.
Neither figure is terribly plausible on its face. As Wired noted earlier this week, 750,000 jobs is fully 8 percent of the current number of unemployed in the United States. And $250 billion is more than the combined 2005 gross domestic revenues of the movie, music, software, and video game industries.
Still, anything is possible: The figures could happen to be more or less accurate. But given the shady provenance of the data, the one thing we know for certain is that we don't know for certain. And we're making policy on the basis of our ignorance.
A 20-year game of Telephone
If you pay any attention to the endless debates over intellectual property policy in the United States, you'll hear two numbers invoked over and over again, like the stuttering chorus of some Philip Glass opera: 750,000 and $200 to $250 billion. The first is the number of U.S. jobs supposedly lost to intellectual property theft; the second is the annual dollar cost of IP infringement to the U.S. economy. These statistics are brandished like a talisman each time Congress is asked to step up enforcement to protect the ever-beleaguered U.S. content industry. And both, as far as an extended investigation by Ars Technica has been able to determine, are utterly bogus.
"I have said it thrice," wrote Lewis Carroll in his poem The Hunting of the Snark, "what I tell you three times is true." And by that standard, the Pythagorean Theorem is but schoolyard gossip compared with our hoary figures. As our colleagues at Wired noted earlier this week, the 750,000 jobs figure can be found cited by the U.S. Department of Commerce, Customs and Border Patrol, and the U.S. Chamber of Commerce, among others. Both feature prominently on TheTrueCosts.org, an industry site devoted to trumpeting the harms of piracy. They're invoked by the deputy director of the U.S. Patent and Trademark Office. And, of course, they're a staple of indignant press releases from the congressional sponsors of tough-on-piracy legislation.
By more conventional standards of empirical verification, however, the numbers fare less well. Try to follow the thread of citations to their source, and you encounter a fractal tangle of recursive reference that resembles nothing so much as the children's game known, in less-PC times, as "Chinese whispers," and these days more often called "Telephone." Usually, the most respectable-sounding authority to cite for the numbers (the FBI for the dollar amount, Customs for the jobs figure) is also the most prevalent—but in each case, that authoritative "source" proves to be a mere waystation on a long and tortuous journey. So what is the secret origin of these ubiquitous statistics? What doomed planet's desperate alien statisticians rocketed them to Kansas? Ars did its best to find the fountainhead. Here's what we discovered.
Looking for lost jobs
First, the estimate of 750,000 jobs lost. (Is that supposed to be per year? A cumulative total over some undefined span? Those who cite the figure seldom say.) Customs is most often given as the source for this, and indeed, you can find press releases from as recently as 2002 giving that figure as a U.S. Customs and Border Patrol estimate. Eureka! But when we contacted CBP to determine how they had arrived at that imposing figure, we were informed that it was, in essence, a goof. The figure, Customs assured us, came from somewhere else, and was mistakenly described as the agency's own. This should come as no great surprise: CBP is an enforcement agency, whereas calculating the total loss of jobs from IP infringement would require some terrifyingly complex counterfactual modeling by trained economists. Similar claims have appeared in Customs releases dating back at least to 1993, but a CBP spokesperson assured us that the agency has never been in the business of developing such estimates in-house.
With Customs a dead end, we dove into press archives, hoping to find the earliest public mention of the elusive 750,000 jobs number. And we found it in—this is not a typo—1986. Yes, back in the days when "Papa Don't Preach" and "You Give Love a Bad Name" topped the charts, The Christian Science Monitor quoted then-Commerce Secretary Malcom Baldridge, trumpeting Ronald Reagan's own precursor to the recently passed PRO-IP bill. Baldridge estimated the number of jobs lost to the counterfeiting of U.S. goods at "anywhere from 130,000 to 750,000."
Where did that preposterously broad range come from? As with the number of licks needed to denude a Tootsie Pop, the world may never know. Ars submitted a Freedom of Information Act request to the Department of Commerce this summer, hoping to uncover the basis of Baldridge's claim—or any other Commerce Department estimates of job losses to piracy—but came up empty. So whatever marvelous proof the late secretary discovered was not to be found in the margins of any document in the government's vaults. But no matter: By 1987, that Brobdignagian statistical span had been reduced, as far as the press were concerned, to "as many as 750,000" jobs. Subsequent reportage dropped the qualifier. The 750,000 figure was still being bandied about this summer in support of the aforementioned PRO-IP bill.
$250 billion? What's that in real money?
What, then, of that $200 to $250 billion range? Often, it's attributed to the Federal Bureau of Investigation, and indeed, the Bureau routinely cites those numbers. According to FBI spokesperson Catherine Milhoan, the figure "was derived through our coordination with industry, trade associations, rights holders, and other law enforcement agencies" at a 2002 anti-piracy confab. But neither the Bureau nor the National Intellectual Property Rights Coordination Center, which assembled the inter-agency powwow, could find any record of how that number was computed.
At this point, it's necessary to get a little speculative. As with Customs, the FBI is not in the habit of doing sophisticated economic analysis in-house. And the last time the government conducted any sort of verifiably rigorous study of the costs of IP theft—about which more presently—it was a protracted undertaking that involved sending detailed questionnaires to hundreds of businesses, which government economists concluded was still insufficient to produce a reliable figure for the economy as a whole. However, $250 billion is about the number you come up with if you start with $200 billion in 1993 dollars and adjust for inflation to 2002. And that lower end of the range, $200 billion, happens to date back to 1993.
Another group that routinely uses the $200 to $250 billion figure is the International Anti-Counterfeiting Coalition, which (along with the FBI) is often given as the source of the number. That organization's white papers, as recently as 2005, footnote the figure to 1995 congressional testimony urging passage of what became the Anticounterfeiting Consumer Protection Act of 1996. So Ars dug into the archives at the Library of Congress to discover where the witnesses before the House and Senate Judiciary Committees got their data.
Several of the witnesses were conspicuously vague about their sources. An IACC factsheet submitted for the hearings said the group itself "estimates the economic cost due to product counterfeiting to exceed $200 billion each year," a number repeated by the group's then-president, John Bliss. Congressman Bob Goodlatte (R-VA) gave the same figure without sourcing. But several witnesses pointed to Forbes magazine as the source of the number. Rep. John Conyers (D-MI) noted that the International Trade Commission had placed the size of the counterfeit market at $60 billion in 1988 and that "a more recent estimate by Forbes Magazine says that American businesses are losing over $200 billion each year as a result of illegal counterfeiting." Finally, Charlotte Simmons-Gill of the International Trademark Association was kind enough to give a precise citation: the October 25, 1993 issue of Forbes.
Ars eagerly hunted down that issue and found a short article on counterfeiting, in which the reader is informed that "counterfeit merchandise" is "a $200 billion enterprise worldwide and growing faster than many of the industries it's preying on." No further source is given.
Quite possibly, the authors of the article called up an industry group like the IACC and got a ballpark guess. At any rate, there is nothing to indicate that Forbes itself had produced the estimate, Mr. Conyers' assertion notwithstanding. What is very clear, however, is that even assuming the figure is accurate, it is not an estimate of the cost to the U.S. economy of IP piracy. It's an estimate of the size of the entire global market in counterfeit goods. Despite the efforts of several witnesses to equate them, it is plainly not on par with the earlier calculation by the ITC that many had also cited.
But here, at last, we have a solid number to sink our claws into, right? Sure, it's 20 years old, but the U.S. International Trade Commission at least produced a reputable study yielding a definite figure for the cost of piracy to the U.S. economy: $60 billion annually.
Well, not quite.
"Biased & self-serving"
The number the ITC actually came up with, based on a survey of several hundred business selected for their likely reliance on IP for revenue, was $23.8 billion—the estimated losses to their respondents. That number was based on industry estimates that the authors of the study noted "could admittedly be biased and self-serving," since the firms had every incentive to paint the situation in the most dire terms as a means of spurring government action. But the figures at least appeared to be consistent and reasonable, both internally and across sectors.
The $60 billion number comes from a two-page appendix, in which the authors note that it's impossible to extrapolate from a self-selecting group of IP-heavy respondents to the economy as a whole. But taking a wild stab and assuming that firms outside their sample experienced losses totaling a quarter to half those of their respondents, the ITC guessed that the aggregate losses to the economy might be on the order of "$43 billion to $61 billion."
The survey also, incidentally, asked respondents to estimate the number of job losses they could attribute to inadequate intellectual property protection. The number they came up with was 5,374. If we assume, very crudely, that job losses are proportionate to dollar losses, then the ITC's high-end estimate of $61 billion in total economic costs would correspond to a loss of not 750,000 jobs, but 13,774.
If we want to be very precise, however, we should note that the ITC was not calculating losses from IP "theft," but rather "inadequate protection" of intellectual property. And "inadequate protection" was interpreted to mean protection falling short of the level provided by U.S. law. The protection provided by a foreign country might be deemed "inadequate," the study explained, if "exceptions to exclusive rights are overly broad"—for example, if a country's law contained "broad exceptions for public performances in hotels or film clips" or "too broad exceptions for educational photocopying." A legal regime could be "inadequate" because "terms of protection are too short" or because of "inadequate" civil or criminal remedies, meaning monetary damages or criminal penalties for infringers were not high enough.
Calculating the net cost of piracy to the economy
One final, slightly theoretical point deserves emphasis here. All the projections we've discussed, the rigorous and the suspect alike, calculate losses in sales or royalties to U.S. firms. This is often conflated with the net "cost to the U.S. economy." But those numbers—whatever they might be—are almost certainly not the same. When someone torrents a $12 album that they would have otherwise purchased, the record industry loses $12, to be sure. But that doesn't mean that $12 has magically vanished from the economy. On the contrary: someone has gotten the value of the album and still has $12 to spend somewhere else.
In economic jargon, charging anything for pure IP—which has a marginal cost approaching zero once it has been produced—creates a deadweight economic loss, at least in static terms. The actual net loss of IP infringement is an allocative loss that only appears in a dynamic analysis. Simply put, when people pirate IP, the market is not accurately signaling how highly people value the effort that was put into creating it, which leads to underproduction of new IP. To calculate the net loss to the economy over the long run, you'd need to figure out the value of the lost innovation in which IP owners would have invested the marginal dollar lost to piracy, and subtract from that the value of the second-best allocation—which is to say, whatever the consumer of the pirated good spent his money on instead—and the value of the deadweight loss (free music or software is a net economic benefit to someone) incurred by pricing IP at all.
If that sounds incredibly complicated, it is. And in fact, it's more complicated than that, because as Yochai Benkler has argued persuasively, IP is an input to innovation as well as the product of innovation. So under certain very specfic conditions, "piracy" can produce net gains. While it seems extremely unlikely that this is the case in the aggregate—IP theft almost certainly does impose net economic costs—simply calculating lost sales and licencing fees, assuming someone could produce a credible figure, would not provide a complete picture of the economic impact of IP infringement. It would give us, at most, one side of the ledger.
Conclusions
But enough theory and speculation; here is what we can say for certain: the two numbers that are invariably invoked whenever Congress considers the need for more stringent IP enforcement are, at best, highly dubious. They are phantoms. We have no good reason to think that either is remotely reliable.
Perhaps more importantly, both numbers are seemingly decades old, gaining a patina of currency and credibility by virtue of having been laundered through a relay race of respectable sources, even as their origin recedes into the mists. That's especially significant, because these numbers are always invoked as proof that the piracy problem is still dire—that everything we've done to step up international enforcement of intellectual property laws has been in vain. But of course, if you simply recycle the same numbers from 15 and 20 years ago—remember that IACC's 2005 publications still cite that 1995 congressional testimony, from which it seems safe to infer that they have no more recent source—then it will necessarily seem as though no ground has been gained.
Neither figure is terribly plausible on its face. As Wired noted earlier this week, 750,000 jobs is fully 8 percent of the current number of unemployed in the United States. And $250 billion is more than the combined 2005 gross domestic revenues of the movie, music, software, and video game industries.
Still, anything is possible: The figures could happen to be more or less accurate. But given the shady provenance of the data, the one thing we know for certain is that we don't know for certain. And we're making policy on the basis of our ignorance.
No comments:
Post a Comment