Chapter 5: Bubbles and Storms: The Story Behind the Numbers

Aram

June 4, 2012

¶ 1 Leave a comment on paragraph 1 0 We are all familiar with this story: Everything was going swimmingly for the music industry until Napster hit. Sales were on the rise, and the future looked brighter still. But since that fateful day in the Summer of 1999 when P2P file sharing was unleashed upon the world, music sales have plummeted and a once-vital industry has been reduced to a shadow of its former self. As RIAA chief executive Cary Sherman lamented in The New York Times in 2012, “music sales in the United States are less than half of what they were in 1999, when the file-sharing site Napster emerged, and [as a result] direct employment in the industry ha[s] fallen by more than half since then.”[1]

¶ 2 Leave a comment on paragraph 2 0 That P2P is squarely to blame for this turn of events is rarely questioned. The recording industry maintains that “[w]idespread piracy is the biggest factor undermining the growth of the digital music business,” and continues to push for “cooperation from online intermediaries” such as ISPs and search engines (largely in the form of surveillance and censorship) as a remedy, or at least a bulwark, against the tide of P2P and other “unauthorized channels” of music distribution.[2] Stanley Liebowitz, an economics professor whose research on file sharing has been funded[3] – and often cited[4] – by the RIAA, even claims that “file-sharing has caused the entire enormous decline in record sales that has occurred over the last decade.”[5] The news media tend to reproduce this frame of analysis without critique, routinely referencing “losses from file sharing” or speaking of sectors “avoiding what happened to the music industry” in their business coverage.

¶ 3 Leave a comment on paragraph 3 0 If this narrative has succeeded in becoming “common knowledge,” a truism repeated in classrooms, boardrooms, and cocktail parties around the world, it has been aided in large part by the Chart. This graphical argument has appeared in various forms, in hundreds of blogs and publications, but each version tells essentially the same story: a market peak, followed by the introduction of P2P, followed by a long and steep decline. An excellent example is the version of the Chart provided by Liebowitz in his testimony on behalf of the plaintiffs in Arista v. Lime Group (Figure 3), which has been reproduced in The Hollywood Reporter[6] and elsewhere. As Liebowitz’s chart shows, music sales in the US, measured in terms of albums sold per capita, did indeed reach a historical market peak shortly before the introduction of Napster, and have fallen significantly since then. He also asserts that music sales would have continued to climb linearly, without leveling off or falling, had P2P not undermined this growing consumer demand (a claim that seems to defy the basic tenets of logic). As he argued in Sony BMG v. Tenenbaum, another file sharing case in which he was retained as an expert witness by the major label plaintiffs, “the clearest and probably the most compelling evidence for file-sharing’s impact on sound recording sales is the timing of the rise of file-sharing with the decline in sound recording sales.”[7] In other words, according to the music industry, the correlation between these two events is the greatest proof that the former caused the latter.

¶ 4 Leave a comment on paragraph 4 0  

¶ 5 Leave a comment on paragraph 5 0

¶ 6 Leave a comment on paragraph 6 0 Figure 3: Chart Depicting The Effect of P2P on Music Sales by Stanley Liebowitz

¶ 7 Leave a comment on paragraph 7 0  

¶ 8 Leave a comment on paragraph 8 0 As I will argue in this chapter, the Chart and its accompanying narrative, though they contain elements of truth, amount to little more than a convenient fiction, scapegoating music fans and media innovators for the recording industry’s own strategic failures and ascribing responsibility to “pirates” and “piracy” for trends and events that are beyond anyone’s control. While the introduction of Napster does correlate conveniently with the beginning of a downward trend for music retail, so do a number of other factors, and furthermore, as any statistician can tell you, correlation doesn’t necessarily imply causation. As one statistics textbook puts it, “an observed correlation between two variables may be spurious. That is, it may be caused by the influence of a third variable.”[8] Below, I will describe many other variables that have played a role in the transformation of the music economy over the past few decades, demonstrating that, to the extent that P2P plays any part, it is relatively minimal. The larger story involves a “perfect bubble” – a confluence of economic, political and technological forces that drove the aggregate value of music sales to unprecedented heights at the end of the 20^th Century – followed by a “perfect storm,” which punctured this bubble and undermined the music retail market. I will also briefly deal with the music industry’s often-cited figures regarding the economic impact of piracy on jobs and productivity, showing that independent research has debunked many of these claims.

¶ 9 Leave a comment on paragraph 9 0 Before we discuss these points, however, I will offer a chart of my own, depicting the IFPI’s own published figures for the global music sales market (Figure 4). It parallels Liebowitz’s in many respects, although it represents actual money spent on music rather than unit sales per capita. Most saliently, there is a steep two-decade climb, followed by a peak around the turn of the century, followed by a steep decade-long dip. Also like Liebowitz’s and every other version of the Chart, this is as much a work of art as a work of science. All data and methods of analysis have their biases and inconsistencies, and market research published by the music industry excels in both of these respects. Therefore, any researcher working with the data must necessarily use his or her judgment in developing a meaningful set of figures as a basis of analysis. (If you aren’t keenly interested in the detailed challenges of working with music industry market data, feel free to skip to the next section).

¶ 10 Leave a comment on paragraph 10 0 There is no definitive tally of music industry market data. The IFPI, the RIAA and other official organs of the music industry regularly publish statistics, and these are often supplemented, cited, and reliant upon data from third-party research companies, such as Nielsen SoundScan. Yet there is rarely agreement even between two publications from the same source, let alone among these many sources. There are a variety of reasons for this. First of all, organizations such as the IFPI and the RIAA routinely revise previously published figures, for a variety of reasons including changes to their internal data, analyses, and methodologies.[9] Second, inflation makes longitudinal data difficult to compare. Some of this is due to confusion (it’s not always clear which year’s dollars are represented in a given publication’s figures), and some of it is genuinely thorny math (inflation is not consistent from region to region, market to market, and currency to currency). A separate but related challenge is the fact that exchange rates between currencies differ on a daily basis; therefore, globally-reported market figures in US dollars are difficult to assess for a single year, and a guesstimate at best for longitudinal data. As the Organisation for Economic Co-operation and Development (OECD) warned in its own analysis of IFPI data, “Global sales figures in USD . . . must be used with caution due to fluctuating US dollar exchange rate which can make year-to-year comparisons difficult.”[10]

¶ 11 Leave a comment on paragraph 11 0 In addition to these macroeconomic challenges, many of which pertain to any global marketplace, there are additional idiosyncrasies about recording industry data that make them even knottier to unravel. For one thing, there is a methodological inconsistency between some figures, which are based on analysis of retail sales data (e.g. SoundScan, which the IFPI uses) and others, which are based on shipments, or the number of units record labels report sending to retailers, which are then extrapolated to dollar figures (e.g. RIAA publications). Each of these methods of assessment has its strengths and weaknesses, but there is invariably a significant quantitative gap between the two.

¶ 12 Leave a comment on paragraph 12 0 Another challenge is that the music industry sometimes reports sales data in terms of “trade value,” or wholesale price, and at other times reports data in terms of retail value, or the price paid at market. Moreover, the conversion rate between retail and trade value differs from format to format, region to region, and year to year. For instance, the IFPI recently reported figures that suggested an 83% retail markup for physical goods and a 59% markup for digital goods in the US in 2008, while reporting a 107% markup for physical and a 73% markup for digital in Austria in the same year.[11] Given that the IFPI’s older publications report retail value, while the newer ones tend to report trade value, this makes longitudinal market analysis even more difficult.

¶ 13 Leave a comment on paragraph 13 0 Finally, there is the question of what the object of analysis is. Historically, the music industry only reported revenues accruing from the global sale of physical goods in brick-and-mortar stores. As the industry’s revenue model has diversified, some additional income sources have slowly been added, though others have not. For instance, global IFPI figures have included performance royalties and digital sales (including ringtones) for most of the past decade, and in 2012 began to include synch license royalties. Given that these aren’t technically “sales,” and have no retail markup, the process of comparing current to past global market figures is a bit of an apples-to-oranges-to-watermelons process.

¶ 14 Leave a comment on paragraph 14 0 I discuss these challenges not to bemoan my job as a researcher or to besmirch the integrity of the recording industry, but simply to point out that it is theoretically impossible to describe the historical global music marketplace with total accuracy, and that any market data that appear in any publication must be understood as fundamentally interpretive in nature. Nor are the resulting inconsistencies sufficiently small as to be of academic interest only; they bear directly on the questions I address in this chapter: namely, when and why did the music industry’s fortunes reverse? According to an IFPI publication in 2000, the global music retail market peaked in 1996, followed by a market contraction, with a 2% drop in 1999.[12] By 2005, the IFPI was reporting two sets of figures: in terms of “variable” dollars (at same-year exchange rates), the market peaked in 1996, but in terms of “fixed” dollars (all years calculated at a 2004 exchange rate), the peak came in 1999.[13] Today, all IFPI publications show a 1999 peak, and there is no discussion of fixed vs. variable dollars. Perhaps the industry changed its analytical methods because it believes fixed dollars are a more meaningful measure. Perhaps it just makes for a better story.

¶ 15 Leave a comment on paragraph 15 0 For my own version of the chart (Figure 4), I rely upon data from two recent IFPI publications. For 1969-2004, I use the variable-dollar figures reported in The Recording Industry in Numbers 2005. I chose these figures because they are the most recent official data going back that far, because they reflect retail rather than wholesale, and because variable dollars more accurately reflect the role of macroeconomic factors (such as buying power) in shaping the music economy over time. For more recent years, I use the IFPI’s numbers published in 2012 (which presumably use fixed dollars). Because these are reported in terms of trade revenue, I adjusted for retail based on an average 70% markup across different regions and formats. As I discussed above, the IFPI doesn’t use a single conversion rate, but this figure is both conservative relative to the range of percentages the industry uses, and consistent in its results with many additional published market data.[14]  All of my figures reflect inflation-adjusted 2011 US dollars.