Rockonomics

by Alan B Krueger

  This calculation assumes a lot of good faith. Some streaming services have been accused of inflating streams for favored artists or those with an equity stake in the company. And questions linger about how revenues are computed and whether revenues are shifted to other aspects of the business to avoid being subject to royalties. The streaming service Tidal, for example, is under investigation for its business practices.11 Considering that many record companies used to underreport album sales in the 1960s and concert promoters used to underreport ticket sales and exaggerate expenses in the 1970s, I would not be surprised to learn that some streaming companies have used creative accounting practices to boost their profits and overstate their reach. The history of rockonomics teaches us to trust but verify.

  This analysis oversimplifies the revenue issues, however. In practice, more is negotiated between the streaming platforms and record labels than just the payout rate. A confidential and murky feature of the streaming business is that record labels sometimes implicitly compensate streaming platforms to undertake campaigns to promote their artists. Promotional campaigns are sometimes obvious. For example, Spotify pushed Drake’s hit album Scorpion so pervasively that many fans complained on social media. Scorpion was featured in multiple Spotify playlists, and Drake’s picture was displayed on playlists that were unrelated to his music, such as “Best of British,” “Massive Dance Hits,” and “Happy Pop Hits.” Apple Music also heavily promoted the album, creating a website where users could create their own Drake album cover art and prepping Siri with answers about Drake. This effort was highly successful: Scorpion broke Spotify’s one-week U.S. streaming record in its first three days, and Apple’s single-day streaming record.12 And Scorpion occupied the number-one slot on the Billboard Top 200 chart longer than any other album in 2018.

  A less well-known practice is that labels and platforms can also tie promotional efforts for new artists to their deals with streaming services. And Spotify has experimented with paid sponsored songs on its ad-supported service.13 These marketing efforts raise questions of whether negotiated promotional tie-ins cross a line into payola, the now-illegal practice of record labels paying radio stations to play their artists’ music.

  Streaming services compete with one another on a number of dimensions. Although the major platforms all have a similar catalog of music available for premium subscribers, they design different playlists and have different features, such as voice interaction. Unlike broadcast radio, streaming services collect voluminous information on the history of individual listeners’ music choices and use this information to make personalized listening recommendations. And the free-form nature of voice-interactive services enables Amazon Echo Dot and Google Home to collect different types of information (e.g., users’ interest in dinner music) than the more regimented screen-based streaming services; they can use this information to create customized playlists and recommendations for users. Furthermore, streaming services invest in metadata to code different features of songs, to better match music selections with listeners’ preferences.

  One aspect of such personalized service is that information becomes an asset for streaming services. This asset creates “individual-specific match capital,” in that a streaming service is able to provide a more customized service for its clients than a rival streaming service could without that listener’s history. The better that platforms become at delivering personalized services, the less likely subscribers will be to move to another platform, and the more scope streaming services will have to raise prices for long-standing customers.

  Some streaming platforms provide complementary services and have created complementary products from their activities. For example, Spotify offers recording artists a dashboard of data on their streaming performance, which they can use to draw insights about their fans, target audiences, and select touring locations. Apple provides a similar service, and Amazon has promised to follow suit. And Spotify together with Songkick provides information about concerts based on listeners’ interests. Some services also have links to merchandise for sale.

  Streaming platforms gain many advantages from scale. They are in a stronger position to bargain with labels if they are larger, and they can provide better service to listeners and artists if they have data on a larger number of users’ preferences. And even though the platforms pay a more or less constant share of revenue in royalties, some costs are fixed (e.g., developing a web platform and recommendation algorithm) and therefore do not rise as more customers are added. As a result, there is intense competition among the various streaming services to grow and expand to new regions of the world, such as India and Latin America. Just as scale and uniqueness create superstars in music, conditions are ripe for a handful of streaming companies to dominate the music market in the future.

  A final aspect of the business model worth noting is that streaming services are beginning to price-discriminate to maximize revenue and increase their number of users. Although the most common subscription price is $9.99 a month, prices vary depending on class of service and number of devices or family membership. Amazon currently offers its Prime members a limited catalog (two million songs) for free and an unlimited catalog for $7.99 a month ($3.99 a month to use the service specifically on one Echo Dot device), and a family plan for $14.99 a month where the unlimited category can be streamed by up to six family members. Spotify has a free on-demand service with ad interruptions and a premium individual and family subscription plan without ads. Apple Music does not offer a free service but has an individual plan for $9.99 a month and a family plan for $15 a month for up to six family members. YouTube, which is overwhelmingly an ad-supported platform, recently started to offer ad-free music for $11.99 a month and a family plan for $17.99 a month.14 Many platforms offer lower prices to students. The goal of these pricing strategies is to charge a higher price to consumers who are less price sensitive, in order to increase revenue and increase market share.

  The Economics of Streaming

  A useful approach for understanding the economic impacts of any innovation is to consider the effect of the innovation on the size of the pie (i.e., whether the total amount of income rises or falls) and on the distribution of the pie (i.e., how large a slice of the pie various parties get). The advent of paid streaming services has undoubtedly increased the size of the pie. More money is being spent on recorded music because of streaming over the last few years, and this is to the benefit of musicians and labels as a whole. The anticipated growth in the number of streaming subscribers should increase the size of the pie further in the future.

  The distribution of the pie is more complicated, and depends on the contracts that particular artists have with their labels. Nonetheless, it is clear that some genres have benefited more from streaming than others. For example, hip-hop and R&B represented 15.5 percent of total album sales and 36 percent of total on-demand music streams in the first half of 2018, according to Nielsen.15 Dance and electronic music also represented relatively more on-demand streams than album sales. The fact that hip-hop, R&B, and EDM outperform their album sales on streaming services is largely a reflection of the relatively young and urban customer base that is using streaming services.

  Streaming is also giving a second wind to older songs, known in the trade as the back catalog. Publishing and performance royalties are an annuity, and a vibrant financial market has developed to securitize and sell shares in artists’ royalties. This development was foreshadowed by David Bowie, who securitized the royalties from twenty-five of his albums to raise $55 million in eponymous Bowie Bonds in 1997.

  Many new artists, however, may receive a smaller slice of the pie with streaming compared with physical records or digital downloads, at least initially. The reason is that when music was a durable good that was purchased and then listened to later (often multiple times), new releases were more likely to be purchased than older releases. Emerging artists, who tend to relea
se more new music, therefore are initially receiving less revenue with the transition to streaming. It is possible that the size of the pie will grow large enough that new artists are eventually better off with streaming as well, if one considers the present value of royalty payments under streaming compared with the previous state of the record industry.

  Three Misconceptions

  Because streaming is a new product with a different business model, there are three mistakes that people commonly make concerning the economics of streaming.

  First, it is common for industry insiders to view streaming as a zero-sum game. The logic is that if one artist’s streams rise, that artist’s share of total streams will rise, and the share earned by other artists’ will necessarily fall. In other words, instead of vying individually to sell more records, artists are now viewed as competing against one another to earn a larger share of the pie. This logic holds in a static world but not in a dynamic one, as the number of paying customers can increase the size of the pie over time. And even in a static world, the size of the pie is so large that the quantity of one artist’s streams does not meaningfully affect the share accruing to any other artist.

  Consider the following calculation. With 2.3 billion streams across audio platforms, the Canadian rapper Drake was the most-streamed musician in the world in the first half of 2018.16 His streams accounted for 1.7 percent of all songs streamed in that period. If Drake had decided to become a lawyer instead of a musician and his catalog of songs was never produced—and nothing else had changed—by how much would Kendrick Lamar’s share of streams have increased? The answer is: not much. If we remove Drake’s streams from the total, Kendrick Lamar’s share of streams would only increase from 0.50 percent to 0.51 percent. The benefit for Cardi B is even smaller, as is the case for every other artist who was streamed less frequently than Kendrick Lamar, given the nonlinear effect of shaving the denominator. And removing artists with fewer streams than Drake would have even less effect on other artists’ shares. Thus as a practical matter, even in a static view of the world, there is little zero-sum aspect to competition among artists with streaming.

  Furthermore, the fact that streaming is not yet close to reaching its maximum number of paying customers is another reason streaming is not a zero-sum game; the sum will grow larger if more hit songs cause more customers to sign up. And even when the number of paid subscribers levels off, the size of the pie could still continue to grow if monthly subscription fees increase, or if subscription fees are further tiered to take advantage of price discrimination.

  A second common misperception is the belief that the amount an artist earns per stream from a platform such as Spotify or Tidal is a meaningful indicator of the generosity or contribution of the streaming service to a musician’s income. Music blogs, the New York Times, Forbes, and other news outlets have presented artist payments per stream as a measure of streaming services’ generosity.17 In a subscriber-based streaming market, however, the parameters that determine income from recorded music depend on the platform’s payout rate, the fee that subscribers pay per month, and the number of subscribers. The number of streams is irrelevant. To see this, suppose there are two platforms, A and B, and both charge $9.99 a month. Each has one million paying subscribers, and each pays 70 percent of its revenue to rights holders. Suppose service A’s subscribers stream twice as much music as service B’s subscribers, because service A does a better job of recommending music that subscribers like. In this example, the amount paid per stream would be half as much for service A as for service B. But service A is clearly better for the music industry, because it does a better job recommending music, makes listeners happier, and pays out just as much money to rights holders as service B pays. This example makes clear that the amount paid out per stream can be a misleading indicator of which streaming service is better for artists and the rest of the music community. In fact, all else being equal, more successful services can be expected to pay less per streamed song than less successful ones.

  This is not to say that there is not a “value gap” when it comes to certain streaming services. YouTube is a prime example of the value gap. YouTube, which has only a small subscription service, pays a smaller share of its revenue out to rights holders (as I explain in Chapter 9, this is a result of a quirk of the Digital Millennium Copyright Act of 1998). The other services are fairly similar in their payout rate and monthly fees. Services that rely relatively more on ad-supported customers than on paying subscribers, however, have less revenue to pay out, because ads generate less revenue than subscribers. So the fact that Apple Music pays out more per stream than Spotify may be an indication that Apple subscribers spend less time listening to music than Spotify subscribers (as well as the fact that Spotify is more reliant on ad-supported customers). And the fact that Spotify’s payments per stream have declined over time is a sign that Spotify’s growing number of subscribers are spending more time listening to more songs, which is good for the music industry.

  A final misconception is that there exists a metric to convert the number of streams to album sales equivalents. There is a natural desire to try to measure the popularity of artists or albums, because music is a social good. This has led to a controversial and fraught search for a simple way to combine album sales, digital downloads, and streams into one overarching measure of popularity. Billboard, for example, counts 1,500 song streams from one album as the equivalent of one album sale in its influential Billboard 200 chart.18 But combining a certain number of streams with the number of albums sold to measure popularity is as sensible as using horsepower as a metric to gauge the combined popularity of automobiles and horse and buggies. Streams and album sales are different products, and the yardstick used to combine them is arbitrary. Streaming increases the variety of songs available and reduces the marginal monetary cost of listening to music to zero, so more music is consumed with streaming. Another difference is that albums necessarily bundle songs, while with streaming songs are unbundled (as explained further later).

  From à la Carte to an All-You-Can-Eat Buffet

  From the music consumer’s perspective, streaming converts recorded music from an à la carte menu to an all-you-can-eat buffet that is more convenient than downloading pirated music from unauthorized websites. Once the monthly subscription is paid, the marginal pecuniary cost of consuming music is zero. And as with an all-you-can-eat breakfast, consumers tend to consume more when there is no extra cost of consumption. The main constraint on the consumer’s side is time.

  It has become clear that many consumers are willing to pay for convenience. Spotify was designed to be quicker and easier to use than the free file-sharing services provided by LimeWire, Pirate Bay, and the profusion of other pirate music sites that followed Napster and the BitTorrent wave. The model has worked. In its public filing, Spotify forecasted that it will have ninety-six million paid subscribers, paying a combined $6.5 billion, by the end of 2018.19

  Users tend to sort into a streaming plan that matches their preferences and situation. Individuals with a high opportunity cost of time and more disposable income are more likely to pay for ad-free premium services. Young and low-income listeners are more willing to endure ads that come with “freemium” service. As with lunch, however, there really is no such thing as a free streaming service.

  The Demand for Ad-Supported Streaming

  With primarily ad-supported streaming services such as Pandora or the gigantic Chinese platform QQ, the cost of listening to music is the inconvenience of having to hear (or watch) paid advertisements. Pandora listeners are interrupted by roughly one or two ads every fifteen minutes, amounting to about three minutes of ads over an hour of listening.20 Ads are targeted to the demographics and geography of listeners. Some are entertaining and play to the medium. Nonetheless, ads impose a cost on the listener—they pay with their time spent listening. (Economic tools can be applied even when money does not change
hands.)

  In one of the clearest demonstrations of the existence of a downward-sloping demand curve, Pandora conducted an unprecedented large-scale experiment to measure the extent to which listeners’ demand for music—and for Pandora itself—declines as the volume of ads rises. Between June 2014 and April 2016, nearly 35 million listeners on Pandora radio were subjected to varying amounts of ads.21 The listeners were randomly assigned to one of nine treatment groups, which varied the number of ads and frequency of interruptions that listeners were exposed to over the next twenty-one months. A tenth group formed a control group, which heard the normal number of ads (approximately three minutes per hour). The group with the most ads received more than twice as many ads as the group with the fewest ads. This is an example of the type of A/B experiment that Internet companies like Google and Facebook routinely conduct on unsuspecting users without publicly disclosing their research results. The experiment was so vast that economist David Reiley left academia to lead the Pandora research team. His findings illustrate the power of applying economic analysis to the music business.

  Consumers responded to advertisements the same way they typically respond to higher prices: they demand less of the good (in this case, streams) if ads impose greater nuisance costs. Figure 8.1 documents that over the course of the experiment, time spent listening to Pandora declined for the group exposed to the most ads relative to the control group, while listening increased for the group exposed to the fewest ads. Listening time fell as the number of ads increased because users listened for fewer days or dropped the service altogether. This creates a fundamental trade-off for Pandora and other ad-supported streaming services: they can charge corporate clients higher rates for advertisements if they attract more listeners, but they lose listeners if they increase the number of advertisements. Interestingly, Pandora’s experiment suggested that the company’s revenue could be increased if more ads were delivered. That is, although Pandora would lose some listeners and therefore advertising fees would fall if ads were aired more frequently, this would be more than offset by the extra revenue gained from selling more ads.