History of P2P Filesharing Research

A brief evolution of P2P research The first step in research was to establish any economic harm by file-sharing at all. Then the analyses widened to include comparisons between geographic regions, primary P2P data, the positive effects of file-sharing on sales, a differentiation by popularity of artists, by genre, by age and other demographics of Internet users and finally social welfare and the efficiency of the current copyright regime and possible alternatives to it. Liebowitz (2002) finds that the arguments for Napster as a "potentially serious threat" "remain basically theoretical. ... The evidence that has been put forward to this point does not 57 Aidouni/Latapy/Magnien (2008) have captured almost 9 billion messages involving almost 90 million users and more than 275 million distinct files off an eDonkey server. This data set is currently being analysed by economist Aigrain (2010). 58 See Ipoque 2009: 7 ff. for a comparison of content on eDonkey and on BitTorrent. 59 http://www.bigchampagne.com/. It describes its activities thus: "The BigChampagne Media Measurement BitTorrent monitoring system is comprised of the following: Building a database of active torrents; Creating and maintaining title/metadata databases; Matching the torrent records to the titles/metadata in the databases; Around the clock scraping of seeders and leechers for torrents; Collecting file sizes; Participating directly in relevant swarms; Monitoring downloads directly, performing geographic analysis and more; Reporting and analysing activity at the title (aggregate) level and the individual torrent level for albums, movies, TV shows, etc." (Page/Garland 2008) 11/25 clearly point to the direction of the impact, to say nothing of the magnitude." (22) He does predict that in the not too distant future DRM will end any harm to copyright owners brought about by unauthorized copying. In his (2003) paper there was still no proof of harm. Liebowitz finds that the recording industry's evidence in the 2000 lawsuit against Napster "failed to support any claim of harm," nor was there evidence of any decline in record sales. Nevertheless, he sees "good economic reasons to believe that online file sharing would be harmful to the industry" even though "the evidence to support this claim had not yet surfaced." (2) This was in 2003, the fourth year of P2P file-sharing with participation ranging in the tens of millions,60 the year The Pirate Bay is founded and the iTunes Music Store goes online. Even so, evidence of harm had not yet surfaced. In his 2006 paper, Liebowitz demagogically speaks about "organized file sharing" (10) and tells his readers: "Common sense is, or should be, the handmaiden of economic analysis. When given the choice of free and convenient high-quality copies versus purchased originals, is it really a surprise that a significant number of individuals will choose to substitute the free copy for the purchase?" (24) Nevertheless, he is careful to point out in his conclusion: "We do not yet have enough evidence to draw any but a preliminary conclusion. ... With a technology this young, and markets changing this fast, it would be most unwise to claim too much given the risk that the future may prove a current conclusion to be incorrect." (2006: 24) In 2008 Liebowitz presents his only original research. At a time when the contrivance of using the number of Internet users as proxy for file-sharing has already been widely dismissed, he goes to great length in arguing that it is indeed a valid method. Nevertheless, he seems to be aware of the weakness of the claim of a causality of Internet use and file-sharing. Because he does not have a measurement of file-sharing intensity, he takes the assumption that young people are more likely to file-share than older people to derive a file-sharing "propensity," thus in effect using youth as a proxy. Comparing data on Internet access and age from 99 US-American cities in the years 1998 and 2003 he finds youth to be negatively correlated with record sales. His calculations yield a "reduction in sales due to file-sharing [that] appears to be larger than the actual measured decline in record sales." (29) He explains this as indicating that file-sharing has not only caused the entire decline in record sales but also the failure of an extrapolated growth that would have occurred without it. This is in stark contrast to the recent increases in sales of CDs reported from major markets like Brazil and the UK, that do not correspond to any decrease in file-sharing activity.61 Macro data lends itself to comparisons between cities and countries. Researching the prefile- sharing situation, Hui/Png (2003) look at 28 countries in 1994-98. Zentner (2005, 2006) uses international time-series aggregate data in conjunction with Internet penetration, finding that countries with more broadband-connections have experienced stronger reductions in album sales. Peitz/Waelbroeck (2004) look at 16 countries, Boorstin (2004) and Liebowitz (2008) each at 99 US cities. Oberholzer-Gee/Strumpf (2004/2007), Blackburn (2004) and Tanaka (2004) were the first to use primary P2P data. While Tanaka and Oberholzer-Gee/Strumpf find no effect of downloading on sales, Blackburn finds a negative effect on stars and a positive effect on less known artists. Bhattacharjee et al. (2006) monitored Kazaa for effects of legal actions, showing that while filesharing intensity decreased for a short time, an ample supply of all chart albums remained available for download. That the discovery effect of file-sharing may actually stimulate sales was already assumed by Shapiro/Varian (1999). It was first shown by Blackburn (2004) for unknown artists, in the survey 60 Peitz/Waelbroeck (2004: 76, table 2) give 44.6 million clients for seven P2P protocols active in June 2003, with Kazaa (35 million) being the largest. 61 Tschmuck (2010) lists further empirical "anomalies": Japan, the second most important music market in the world, suffered an 8.2% decline in CD sales pre-Napster between 1997 and 1999, but an increase by 7.9% in 2000. In France CD sales reached a historic high in 2001. The UK-market dropped by 17.7% in 2001, remained at the same level until 2003, rose by 4.4% in 2004 and showed its first strong dip in 2007. These diverse changes are difficult to reconcile with a continuously and globally growing P2P file-sharing population. 12/25 of Tanaka (2004) and in Boorstin (2004) for those older than 24 years. Bounie/Bourreau/Waelbroeck (2005) in their survey find two distinct types of file-sharers: the "pirates" who keep most of their music downloads, substituting them for purchases, and the "explorers" for whom downloading leads to an increase in purchases of CDs. 88% of their respondents obtained free MP3 files. Nearly all of them reported discovering new artists and 70% said that this led them to purchase CDs that they would not have purchased otherwise. "This result illustrates a strong 'sampling effect' among the respondents of the survey." (10) Their two types could thus be rephrased as "music lovers" who download and buy a lot, and "casual music listeners" who download less than half the number of files of the "music lovers" and would not have purchased them. Peitz/Waelbroeck (2006) again observe that music is an experience good characterised by a two-sided asymmetric information problem between sellers and buyers that can be solved by sampling. "The property that sampling allows consumers to find a better match to their tastes, tends to lead to higher profits under file-sharing." (908) Blackburn (2004) was the first to differentiate by popularity of artists, showing that wellknown artists suffer substitution, while unknown artists benefit from a discovery effect. Because the popular artists sell more albums he finds the overall market effect to be negative. For cultural diversity, one can conclude, the effect is positive, as it curbs the crowding-out effect of superstar sales on other albums62 and makes it easier for new and previously unknown artists to break through. Gopal/Bhattacharjee/Sanders (2006) confirm that "as sampling becomes less expensive, the superstar effect is eroded overall, and more users purchase music items based on their actual, not perceived, valuations," (1528) favouring lesser known artists. That also a differentiated look at genre preferences is required for estimating effects on sales and downloads was first shown by Bounie/Bourreau/Waelbroeck (2005). They find that those who download rap music have a significantly higher probability to have reduced CD consumption than those who download pop/rock music (13). Like Zentner (2005) they find a differential effect on international and local repertoire: "Around 8% of [file-sharing] respondents got music from French artists, whereas according to the French recording association (SNEP), French songs represented more than 40% of total CD sales in France in 2003." (8) The survey in Huygen et al. (2009) sheds some more light on file-sharing and genre preference (68 ff.). Boorstin (2004) was the first to show that the impact of file-sharing varies with age. Using Internet access as proxy, he finds that those below 24 years use P2P to substitute music purchases, while those above, because of a sampling effect, complement it with CD purchases. Because the buying power of the older group exceeds that of the younger the overall effect on CD sales is positive. Bayaan (2004) is the first to not simply ask for economic effects on "the music industry" or on "rights-holders" – implying that authors, performing artists, publishers and labels, large and small are one homogeneous actor – but specifically on artists. By not only looking at the gains or losses of firms but of those of other actors as well, he opens up the research perspective to a welfare analysis. He finds that in the best case scenario the "gain for consumers is more than enough to offset the loss of profit incurred by firms and signed artists so society as a whole benefits." (17) Economics traditionally defines social welfare as the sum of consumer surplus and producer surplus. The seminal reference for the welfare implications of unauthorized reproductions is Takeyama (1994). She assumes network effects, i.e. an increase of consumers' valuations of a product with the number of other consumers who adopt the same product. The effect is most obvious for interoperating products like fax machines and computer software, but, writes Takeyama, "there are many less obvious sources of network externalities, including the psychological desire to 'join the bandwagon.'" (155) Earlier studies had reached varying conclusions with regard to the effect of unauthorized copying on social welfare, but did show a consensus that firm profits 62 He points out that "the last album to sell even 7 million copies in one year was ’N Sync’s 'No Strings Attached,' which sold 9.9 million copies in 1999, just as file sharing was born." (Blackburn 2004: 13) 13/25 necessarily decline with copying. In contrast, her paper demonstrates "that, even without indirect appropriation, when demand network externalities are considered, not only can copying lead to greater firm profits, it can produce a Pareto improvement in social welfare." (156) This has been confirmed by Rob/Waldfogel (2006), who IFPI (2010) cites in its support. Asking for consumer valuation of certain albums, they do find that downloading reduces their respondents' per capita expenditures (on hit albums) by 25 US$, but also that it raises their surplus by 70 US$. The reduction of 45 US$ per capita in deadweight loss, i.e. in socially beneficial, but otherwise foregone transactions, is nearly double the reduction in industry revenue (32). Also for Oberholzer-Gee/Strumpf a key question is how social welfare changes with property rights for information goods – de facto, if not de jure – weakened by P2P file-sharing. In (2009), they remind us that "copyright exists to encourage innovation and the creation of new works; in other words to promote social welfare. The question to ask is thus whether the new technology has undermined the incentives to create, market, and distribute entertainment. Sales displacement is a necessary but not a sufficient condition for harm to occur. We also need to know whether income from complementary products offset the decline in income from copyrighted works. And even if income fell, welfare may not suffer if artists do not respond to weaker monetary incentives." (24) Their reading of the empirical research so far shows that none of the three conditions hold. "Consumer access to recordings has vastly improved since the advent of file sharing. Since 2000, the number of recordings produced has more than doubled. In our view, this makes it difficult to argue that weaker copyright protection has had a negative impact on artists’ incentives to be creative." (25) Huygen et al. (2009) in a study commissioned by three Dutch ministries63 looked at the economic and cultural effects of file-sharing in the three sectors of music, film and games. The Netherlands is particularly interesting for two reasons. Its early and wide broadband adoption makes it rank very high in international comparison. By household penetration, average download speed and subscription price in 2008 it was nearly twice as well equipped for file-sharing than the United States. Yet Huygen et al. find that "the number of music downloaders in the Netherlands is slightly higher than the number most recently found in the US ... Whereas the percentage of film sharers in the United States was more or less the same as in the Netherlands between late 2003 and early 2006, the most recent figures [2008] show that the percentage is now substantially higher in the US. The only known figures for the US show that the percentage of game downloaders is also much higher than in the Netherlands." (86 f.) If another proof that Internet penetration is an unsuitable proxy for file-sharing had been needed, this is it. The second reason that makes the Netherlands stand out against most other countries is that its copyright law permits downloading of copyright protected works from file-sharing networks for personal use. Thus the downloading behaviour is largely unaffected by fear of prosecution. The Dutch survey confirmed the existence of two distinct groups of intensive and of casual media users, where the former download and purchase a lot. Music sharers are no less or more likely to be buyers of music than other people but they buy more merchandise and go to concerts significantly more often. Film sharers buy DVDs no less or more often than anyone else, but if they buy, they buy significantly more DVDs than non-file sharers. Game sharers also buy games, and significantly more frequently than non-file sharers (82). While the majority of respondents in the Dutch study report discovering new genres and artists as their reason for file-sharing, 13% of music and film sharers mention "making social contacts." (77) This shows that file-sharing as a cultural practice is significantly different from the use of streaming or all-you-can-eat download services that are often mentioned as legal alternatives. As for the dynamics in the three sectors, Huygen et al. find that only the markets for CDs and for DVD rentals are suffering from a slump. "The markets for DVDs and console games continued to grow impressively after P2P services were introduced, and the cinema market showed sustained growth between 1999 and 2007. The total entertainment market has remained more or less 63 The Ministry of Education, Culture and Science, the Ministry of Economic Affairs and the Ministry of Justice of the Netherlands. 14/25 constant, suggesting budget competition among the various products." (103) Their estimates on the aggregate welfare balance of file-sharing in the Netherlands are strongly positive in the short and long terms, amounting to about 100 million Euro per year. "The gains enjoyed by consumers are more than twice as large as the losses suffered by producers." (107 f.) Their conclusion: "The survey held among Dutch internet users has shown that file sharing is here to stay and that people who download are at the same time important customers of the music industry. ... File sharing and P2P networks have become generally accepted practices and important drivers for innovation. It would therefore be ill advisable to criminalise file sharing by end users. ... The fact that file sharers in the United States buy fewer products may be related to their harsher treatment in that country." (121 f.) Goel et al. (2010) also argue that unauthorized file-sharing is beneficial to public welfare. They point to the deadweight loss that copyright creates by preventing downloads of works that downloaders are unwilling or unable to purchase which "results in lower social benefits without any increase in revenue for media providers." (4) And they point to the monopoly of copyright and the oligopoly that emerged from it in culture industries: "Further, large media companies may historically have stifled creativity by having excessive influence on deciding what types of works get produced and marketed as well as maintained artificially high prices – e.g., by paying radio stations to play certain numbers, selling more expensive albums rather than the single tracks desired by music fans, and promoting more popular artists at the cost of those with niche followings (and smaller potential profits). Lower search, promotion, and distribution costs associated with the Internet may loosen the stranglehold of large companies and promote creativity while providing works that better cater to diverse consumer tastes at competitive prices." (4) Social welfare thus turns out to be a common theme for nearly all studies on the impact of file-sharing. It is a traditional issue when economists look at the trade-off between under-production and under-utilization of public goods, which creative works sans copyright are by nature and which they have de facto become thanks to file-sharing. Moreover, social welfare is of primary concern for public policy makers who cannot look at record label and publisher profits alone, but have to seek to optimize the aggregate surplus of all actors involved: of authors and performing artists, of consumers, of commercial users of copyright works like radio stations and ISPs and of industries that provide complementary goods and services like MP3 players and mobile phones. Policy makers most of all have to consider the current and future welfare of society as a whole, of education, access to knowledge, cultural diversity and innovation. As Benkler has shown so aptly, the special feature of information goods is that they are both the input and output of their own production process. "In order to write today’s academic or news article, I need access to yesterday’s articles and reports. In order to write today’s novel, movie, or song, I need to use and rework existing cultural forms, such as story lines and twists. This characteristic is known to economists as the 'on the shoulders of giants' effect." (Benkler 2006: 9) Preliminary findings Is P2P file-sharing responsible for the slump in recorded music sales or does it create demand? The empirical research literature is inconclusive. If one were to simply add up studies showing a negative effect and those showing no or a positive effect one would find that the two camps are on par. But that is, of course, not a meaningful exercise. What has clearly emerged is that there are quite a number of different dynamics at work yielding a mixed result with respect to album sales, a likely positive result for the music industry as a whole through gains in concert and merchandising revenues, and a clearly positive effect on social welfare through improved market chances for nonstar music, greater cultural diversity and increased consumer surplus. Since it is evident that a mono-causal explanation is not tenable, a number of studies look for alternative explanations. These include the shift in distribution after 1999 with most record 15/25 stores disappearing and trade becoming concentrated in large retail chains such as Walmart, which drive down unit prices and concentrate on bestseller inventory (Oberholzer-Gee/Strumpf 2007: 39), and other media competing for consumer's time and budgets, in particular DVDs, computer games and mobile phones (Oberholzer-Gee/Strumpf ibid.; Huygen et al. 2009: 103). A fairly obvious candidate for replacing a certain share of record sales, the market for used CDs, has not been tested or even mentioned in any of the research papers in our review. The only paper we are aware of that analysed an online market for second-hand cultural goods is Ghose/Smith/Telang (2006) on books. They "speculate that cannibalization may be particularly acute for digital products, such as CDs and DVDs." (17) Tschmuck (2010) finds another plausible explanation in the major label's decision to fade out singles. The single format rose to a high of 800 million units sold world-wide in 1983 after which it nearly disappeared because the profit marging is much higher for albums. Listeners however, dislike albums which they often consider to contain only one or two tracks of interest while the others are dispensable fillers. First Napster and then iTunes re-introduced what consumers really wanted: a large diversity to choose from in single tracks. "Since 2004, when digital sales were reported for the first time, single sales more than quadrupled (!) to 1.5 billion units" in 2008, the same number as that of albums in all formats sold in that year. "It is clear that you cannot earn the same revenue with the same number of single units than with long-play units sold. Therefore the drop in sales is due to the conversion of an album to a single market." Tschmuck cautions against replacing one mono-causal explanation (file-sharing) by another (shift from albums to singles). "However, the 'single market'-thesis contributes a much better explanation for the declining sales in the recording industry than the 'filesharing' thesis." Another factor mentioned by Oberholzer-Gee/Strumpf (2007) is that a period of atypically high sales, when consumers replaced older music formats with CDs, ended at the turn of the millennium. The replacement of formats is illustrated very clearly by Table 2 in Tschmuck (2010), showing the global sales development for different record formats based on IFPI data: The vinyl LP reached its historic high in 1981 with 1,140 million units. By this time the LP's successor, the prerecorded music cassette, was already coming up, reaching its high in 1991 with 1,493 million units. And again by this time the next generation, the CD, had started its rise to its all-time high in 2000 of 2,454 million units, thus more than twice as many units as the LP had sold in its best year. By this time the CD's successor, Internet and mobile downloads were – nowhere in sight. Recall that the Internet turned into a mass-medium after the first web-browser was released in 1993. The Fraunhofer Institute that had developed the MP3 audio compression technology made a software encoder freely available in 1994 and MP3-encoded music started to spread on the Internet, much of it freely offered by independent artists and much of it encoded from commercial CDs and redistributed without authorization. Therefore anybody interested in music had been aware of the powerful features of digital files and was waiting for the logical transition from the CD to the new format to occur. But it did not, until Apple opened its iTunes Music Store in 2003. What did the major record labels do during the intervening ten years? Two things. For one, they tried to suppress the new technology, suing producers of MP3 players and attempting to outlaw the file format altogether. Second, they bet their business on Digital Rights Management (DRM). By means of cryptographic encapsulation it promised an unprecedentedly fine-grained control over copyright works after delivery to the customer. When it became clear that DRM by itself would not be able to fulfil this promise the music industry sought additional legal protection against its circumvention, which was granted in 1996 by the WIPO Copyright Treaty. A number of industry consortia were set up to establish DRM solutions in various sectors of the culture industry. One of the largest was the Secure Digital Music Initiative (SDMI) established in 1998 by the IFPI and joined by more than 200 companies, including all major computer makers, but quietly disbanded in 2001 after failing to reach its ambitious goals. DRM was primarily used for controlling downloads but also CDs. Sony- 16/25 BMG's root kit fiasco of 2005 was likely the last nail in the coffin of DRM (s. Grassmuck 2006). In 2007 all four music majors announced that they would stop using DRM. If one extrapolates the ten year cycle of LP, MC and CD, 2010 should have been the peak of digital download albums. The gaping hole in the fifth column of Tschmuck's table can therefore likely be explained by the music industry refusing to sell their products online without DRM for ten years. Consumers balancing the harm of DRM against the risk of being caught getting a DRM-free version from an illegal download site or, since 1999 from a P2P network, likely chose for the latter. 'Because it's free' the IFPI claims to be the main motivation for file-sharing. But during the same period mobile music downloads turned into a mass market with consumers often paying ridiculously high prices for a few seconds of ringtone of their favourite songs. Thus P2P filled a gap created by industry itself. It is remarkable that none of the studies asking for alternative explanations for the slump in music sales considered the retarding effect of the dead-end street of DRM. Media technological factors are clearly at the centre of the changing socio-cultural practices. With pervasive gigabyte memory in mobile phones, MP3 players, USB sticks, personal computers and online storage services, the CD as information carrier is clearly losing its attractiveness. Filesharing therefore is a symptom rather than the cause of the changing usage patterns. Thus, asking for the effect of file-sharing on CD sales, as the majority of economic studies have done so far, rapidly becomes obsolete. Research on the universe of file-sharing has to look at a much wider scope of dynamics in technology and in social practices.