CITP Symposium: Voluntary Collective Licensing of Music

Dividing the Pot of Gold

By Harlan Yu

Once the licensing fees are collected into the Big Pot, how will the spoils be divided fairly among all the artists? Any proposal will need a mechanism that accurately measures the online popularity of songs and pays artists royalties in an equitable way.

The first question to ask is what we mean by online song popularity. Will the metric be the number of times a song is downloaded or the number of times a song is played? Both seem to have precedents in the existing business model– the “download metric” is analogous to individual CD sales while the “plays metric” is similar to performance royalties. The possibility of streaming also adds complexity, as it can legitimately be considered under either category. The overall metric could be a combination of both downloads and plays, but how this should be done will be the subject of intense debate.

Despite no clear metric to work with, we can still consider how collecting societies would actually go about measuring popularity. There seem to be two broad strategies:

  • Census. A census requires constant monitoring of a user’s downloading or listening habits and produces a near-perfect measurement of the popularity metric that is chosen. For example, iTunes and other music applications could report every single song played back to the collecting societies. Another option would be to work directly within the ISP to monitor all online traffic to detect which songs are being downloaded over P2P services. (I may be wrong, but presumably this is why ISPs are involved in the first place. The role of ISPs should be the topic of another post.) This near-perfect measurement would allow the societies to divide the revenue fairly to all artists, even those on the far end of the long-tail of music consumption.The problem with this is that users lose all privacy about their musical tastes. The “deep packet inspection” option within ISPs could potentially creep into even nastier security and privacy problems.
  • Sampling. One could also sample from various online music services to make an estimate about the relative popularity of songs. It has been suggested that anonymous listening data from services, such as Pandora and last.fm among others, could be mashed together to produce the sample. The difficulty is that there will always be sampling biases and it is difficult to determine whether the drawn sample actually reflects the actual popularity distribution. The big drawback with sampling is that most artists in the long-tail will rarely, if ever, be sampled. No matter how sophisticated the sampling is, there will be these artists who are under-compensated, and some will never be paid at all.

Neither strategy seems ideal. The census privacy problem is an obvious no-go in my book, and I disagree with Fred that sampling should be the answer. While Nielsen can serve as guidance, sampling to set prices to pay a few large media corporations is different than sampling to provide paychecks to thousands of starving artists. Even if the law of large numbers eventually works itself out (i.e. the artist is paid $1000 with 5% probability each month, depending on whether she is sampled), this is no way to live and pay the bills.

So, is there a third way of measuring that is both accurate and privacy-preserving? Perhaps there is a compromise solution, using the market, where users who don’t consider their listening habits to be private can opt-in to the census and pay a cheaper licensing fee. There ought to be other, similar middle-ground strategies out there. Persuading both artists and listeners that a chosen strategy satisfies both requirements will require a high degree of transparency, just as Samantha advocates.

Regardless of measurement strategy, it is intrinsic in the industry’s move, from offline à la carte to an online buffet, that profit will soon come in as one large pie. And when this pie is made out of money, people will cheat to get the biggest slice. With any metric you choose, I can write a program that downloads or plays my favorite artist’s song repeatedly, day and night, 24/7. The best part, of course, is that my favorite artist could be me. You might try to implement technical countermeasures to stop me, but then I will hire a very large botnet to do the same thing and be harder to detect. We’ve seen many arms races like this before and, as it turns out, the security-breakers almost always win (see: all DRM, spam).

Responses to “Dividing the Pot of Gold”

  1. Michael Says:

    This comment loosely mirrors some of the discussion on the “The Slippery Slope from Voluntary to Mandatory” thread, but here goes:

    Consumers don’t particularly care how the pot of gold is divided as long as the artists keep producing music. This fact is behind the prevalence of illegal file sharing in the first place. Only the artist/producer care about the division, and really just want to get as much as possible, more so than getting their fair share.

    Given the problems inherent in determining 1) the actual number of downloads (purchases) and 2) how much the consumer really values that download (price), there isn’t likely to be a satisfactory solution to the proposed problem with any sort of voluntary or mandatory licensing scheme.

    The only viable solution I’ve seen is “pay up front.” In this model, some organization or population pays the artist enough (determined by the artist) to release the music in the first place. After that release, and the content is in the wild, they will probably not get any more money until another release is ready. Repeated production is encouraged and valuation would likely mirror the historic model of income, albeit in discrete intervals and with a lag of one unit, i.e. you get paid more for your next work if your last one was good. The organization could be the government (such as the NEA), corporations/ISPs, individuals, or voluntary public collectives. It is likely this system could devolve into a Medici style patronage system, but that effect might be lessened by the current ease of participation (to some collective) on the internet.

  2. Samantha Murphy Says:

    Paying by number of plays or downloads won’t work online. There are too many ways to fix the numbers. Take Myspace as an example. It’s easy to manipulate your number of friends, plays, downloads etc.

    This is one area where people usually disagree with my radical approach and that is I feel that the pie should be divided equally among all participants, but that rules should be set in terms of how many songs are contributed and on how often of a basis. Yes that means David Bowie gets paid the same amount as an artist you’ve never heard of. In the case of Myspace, perhaps a 50/50 split of the ad revenue between Newscorp and participating artists (those with a CD released).

    This is the direction society seems to want to take it in. If you look at how consumers want to have the tools to express themselves through the music, as well as how fans are every bit as rabid about artists they have relationships with and email every day (such as Jonathan Coulton), it tells us there is a definite leveling of the playing field here. To that end, it makes sense to distribute evenly to all content providers, with certain caveats in place.

  3. Harlan Yu Says:

    Samantha: How does your system control for quality? Could I record myself every time I sing in the shower (and I do sing a lot) and earn more than Bowie? Maybe I’m missing some crucial details here, but this system doesn’t seem like it would work very well online either.

  4. Samantha Murphy Says:

    Harlan:Most artists who aren’t very good don’t have the tenacity to write, record, master and press a CD, not to mention upload it and book tours/shows etc.

    I’m suggesting there be ways to filter out the serious artists from the hobbyists, but in answer to your question on quality, no. There would be no filter for quality, only dedication. Opinions are subjective.

  5. Fred von Lohmann Says:

    I think sampling can be made highly accurate in a digital world, even into the long tail. After all, the collectives (especially if they are competing) will have strong market incentives to promise to reach down into the Long Tail (thereby signing up artists in the tail and keeping the admin fee). Market incentives should drive innovation here, as elsewhere.

    Sampling should ideally come from many different sources and metrics, in order to minimize the risk of cheating that you identify. So some large user collectives (like Last.FM and Pandora) can provide data, as can P2P crawlers (like BigChampagne), as can “voluntary spyware” survey groups (like NPDs), as can survey groups (like Nielsen). P2P apps could even have reporting features built-in (if they want to, not talking mandates here) that users could choose to activate (like TiVo).

  6. joe Says:

    obviously, some of the arguments used in NN would apply here… for example, encryption can thwart packet inspection. It would be interesting to have a Nielsen expert talk about how sampling like this would be different (for example, how do they correct for TV that isn’t actually watched but when the TV is on? This would be similar to downloading a song that you never listen to.)

  7. Jon Healey Says:

    Samantha, I think treating all artists equally in the pool invites a worse abuse than the download/playcount gaming you’re concerned about. It also denies music fans the sense that their enriching the bands they care most about. I suspect that if you asked diners whether they’d like their tips to go to the person who served them or to a community pot for all wait staff, most would prefer the direct route because it’s a great way to provide feedback. Same thing here. If I’m downloading Gang of Four, I want Gang of Four to benefit from that activity.

  8. Samantha Murphy Says:

    I see your point Jon and your tipping analogy is quite appropriate here. However, as a former waitress, it seems pooling tips ends up being a much more fair system at the end of the day. Take into account seating rotation, poor tippers, inconsistency in the kitchen and shifts based on seniority. I can assure you none of us ever wanted to wait on Mel Gibson or 007%, as we all called Pierce Brosnan.

    Fans will always have the option of supporting artists they love directly. More and more artists are implementing digital tip jars and most people know the best way for their money to end up in an artist’s pocket is to go directly to their website. I’m suggesting if there’s a pool of money, the most fair method of distribution taking EVERYTHING into account, is to divide it equally among participants.

    I realize my opinion here is radical and on the surface seems imbalanced, but upon deeper consideration of every facet of the issue at hand, it makes sense. Artists will still have different levels of income outside of this collective (licensing, live show ticket prices, physical CD’s etc) where different variables will come into play.

  9. Harlan Yu Says:

    I’m still considering your idea, Samantha, and am still not convinced. I don’t know how I would even go about defining a reasonable threshold that separates a serious artist from a hobbyist. It’s as difficult as defining art itself. Not everyone values a large number of short songs, or songs with amazing lyrics, or even dedication or effort. Especially with today’s technology, a “hobbyist” (with little dedication) can conceivably put out an album that is more popular than one by a “serious artist.”

    Even if a threshold is set, lots of hobbyists would make the minimum effort to reach the “serious threshold” just to qualify for the paycheck, and thus lowering the amount everyone gets. Also, how would you ever get a “big” artist to agree to join this kind of collecting society, or stop artists who “get big” from leaving it? It doesn’t seem like the incentives are in the right place on either end of the distribution.

    Assuming something like this existed today, do you have a sense of how big these payouts would be? (Probably look at current data from PROs, but I don’t have this data.)

  10. Crosbie Fitch Says:

    Samantha, we are all artists and we all publish.

    We could pay $5 each into a pot each month, and then this pays us each $5 back - given we’re all publishing artists. Now if one of us happened not to be an artist, then they wouldn’t get $5 back, but that $5 would be split into a million fractions of a cent, and one day it might amount to something worth paying each of the rest of us, say 1 cent.

    What would actually happen is that from the pot, the administrators would take a 40% admin fee. That would leave everyone else with $3, thus everyone is out of pocket to the tune of $2 just for paying an administrator.

    So, why don’t we simply not put any money into the pot in the first place?

    It would be far better if we simply left everyone to pay a commission to each of their favourite artists.

  11. Michael Says:

    Responding to Samantha’s comment about tipping: I recently heard more about the practice of pooling tips (regarding the Starbucks case), and I now tip less as a result of it. Given that my tip is no longer only rewarding good service, but mediocre service in general, it seems less like a tip and more like a fee or surcharge. I suspect I’m not the only one who has made this small, perhaps subconscious change.

    Given the high proportion of low quality music and art produced, I would fight every attempt to pay for music if I were paying mostly for music I don’t want.

  12. Matt Earp Says:

    Hi all. I haven’t piped in yet but this is a thread that’s close to what I’m working on for my thesis. Before going into it, I should say that not only am I a master’s student interested in this stuff, I’m also DJ and music journalist who considers himself a musician and an artist even though I’ve never have created content, at least not in the same manner that Samantha has. But all I do is hang out with musicians who use technology (and who are all the biggest file-sharers I know), so I feel OK speaking up for us artists for some degree. But I have pretty jumbled perspective.

    But I did get into the whole idea of VCL because I saw it as a way for artists to get paid based on the existing listening/consumption/curiosity practices of fans, plain and simple. In the obscure end of the electronic music world the I reside, if you’re talking about moving physical units (records, in this case), 500 records is doing great, 1000 is awesome, 5000 is like you’ve made the new Thriller. But file-sharing is a huge part of the community as well (as are free give-aways and free digital labels). The Soulseek network specifically (which never gets mentioned in the larger debates). I never want to say “file-sharing has cut into the artists I listen to’s profits” because out at my end of the long tail the promotion that comes from giving away free music is appreciable, and the idea of going to court is laughable. Still, there’s no doubt that file-sharing has influenced business decisions, and while peer-2-peer might skim something off the top of a Britney Spears release, it can decimate a small but popular release by one of my friends in no time.

    All that as a lead up to saying: I actually favor the Census style model of tracking, with some caveats. The technology exists to do it in limited settings, like the test case I’m theorizing at Berkeley. Such accounting takes full advantage of the potential of that technology on behalf of artists. I still worry that artists very far out in the tail would be missed by sampling metrics… if, under the new system, an artist has a track that’s downloaded only once a year across the entire web, why not make the best attempt to account for that one download and get them their 1 cent (or probably more like .0001 cent, but still)…. if Itunes can achieve that level of accuracy, at least in accounting, can a VCL system really do any less? Given all that, why not use it to track and payout every artist who wanted to include music in the system, for every single track?

    There are, of course, caveats:

    1) Privacy is a major concern. My short-term solution to that is not to record information on WHO downloads, only on what is downloaded. You finish downloading one of the OK Go’s tracks to your computer? The accounting for that song goes from 500 “hits” to 501. And only that information is collected. Now, systems are only as good as those who maintain them, and I did work at the EFF during some of the AT&T case, so I’m no stranger to large corporations collecting more information than they say they are. A such, that may be another “slippery slope.” But in theory I think it could be tried, with strong privacy experts involved in system design.

    2) Gaming. When this comes up, specifically around bots inflating numbers of a single artist, I say “Aren’t we talking about one of the most fabulously corrupt industries on Earth here? The one that invented payola and never really let it go? Who’s whole history is built on crazy numbers that no one really understands?” To me, I’m much more worried backroom deals and subtle influences by majors on sample-type schemes that has the potential to be much less transparent. Even if it’s not as specific, if it’s just the major labels, the RIAA, and Big Champaign (the only real player in the space) in a room hashing out sample metrics, I worry.

    If a Census-style system is to be set up, it should do what it is intended to do and be as unbreakable as possible (here’s where I’m hoping CITP would have lots to say). In order to prevent gaming, I’d marry census-type data collection with occasional but regular sample based checks. It is an imperfect solution, but to me it’s a start.

    Finally, Samantha, I don’t think it’s crazy at all to suggest ways pool resources, and I applaud thinking that way. It’s not even without precedent. Much of the way rock music was performed and produced in the Soviet Union in the 70s and 80s was through “clubs” where everyone, including the musicians(!), paid a small monthly fee regularly into a pot, and that was divided between among the musicians that were part of the thriving club based on some metrics. Or so I’ve been told by ethnomusicologist friends. But therein lies the rub, my example is Soviet. I think such ideas have a long uphill battle in capitalist society, where people already think of VCL as “socialist”…

    lots of thoughts there

  13. Michael Says:

    The notion that all music or art has equal value is ridiculous at face value. I am a little surprised that several people seem to be supporting it.

    It would seem that the only way to determine fair value for music is to allow the listener to decide. I might value one Britney Spears song as much as ten compositions by Howard Shore (there is some utility in absurd comparisons), so I would want Britney to get not 1/11th, nor equal, but more of the pool.

    Here’s a simple suggestion: Let the consumers police themselves. If your favorite artist no longer gets enough money to keep working, convince other people to pay more or do so yourself. As a producer, make art good enough that people will want to pay you to keep producing it. If they really value it they will pay, perhaps by voluntarily forming organizations or groups to pool their money. Leave it to the consumers to deal with the leeches such as file sharers. It is as simple as that.

  14. Jon Healey Says:

    Well, as long as we’re putting our credentials on the table, Matt, I’m a former recording artist with one released LP to my credit, back in ‘87. Vinyl only. My band made all of $90 in royalties on it, which should explain why I don’t like the sampling approach. Ninety dollars wasn’t much, but it’s certainly more than we would have gotten under a sampling regime. The goal here should be to support the no-barriers-to-entry nature of music in the Internet Era, not to create a threshold for compensation.

    I’d vote for a census approach, but have ISPs do the tracking. I’m not worked up about the privacy implications of that. I don’t think it’s too much to ask consumers to let their ISP identify what they download for the sake of paying artists and labels — especially given that they’re free to download as much as they like without worrying about being sued (provided that the ISP found a way to block content from artists that opted out of the blanket license). To guard against profiles being developed of users based on their downloads, you could limit how ISPs use and retain the data. For the sake of combating fraud, though, I think it’s important to pay attention to IP addresses so that no more than a few downloads or streams of any title would be counted per address.

    I’m not sure that content-identification technology is up to the task of identifying all the song files passing through an ISP’s network. But the job may prove much easier in an environment where file-sharing is legal.

    By not doing the census at the user level, you’d give up the ability to count how many times a song is played. That’s fine by me. True, it would treat a song that someone downloaded and played 1,000 times the same as one that was downloaded but never played. But that doesn’t matter in the CD world — the price of a disc doesn’t go up with repeated listenings — so why should it matter for downloaded works?

  15. Fred von Lohmann Says:

    Can anyone tell me whether a census system administered at the ISP level is actually technically possible?

    Because I am somewhat skeptical. After all, even today we have a constantly changing hodgepodge of protocols — gnutella, bit torrent, mp3 over http, flash over http, fast track, http://ftp. And tomorrow, a new protocol might be invented and become popular overnight. Can an ISP hope to deploy tools that can do a census accounting of all that?

    And that’s not even mentioning encryption, which appears to be driven today by ISP interference for network management reasons (Comcast v. Bit Torrent), rather than copyright. After all, if encrypted traffic is more *reliable* than nonencrypted traffic, because it evades network management, clients will encrypt by default and end-users won’t change the defaults…

  16. Harlan Yu Says:

    Jon: I’m concerned about the census approach because of the privacy precedents it would set. The goal to pay artists fairly is obviously good and undisputed, but there should be few good reasons to allow anyone to monitor traffic at the ISP level. If we allow the music industry to begin monitoring, other IP rights holders will soon be clamoring to get in on the action too. So will online behavioral marketers, who are already knocking on the door. They will all surely promise to protect user privacy, and we’ve seen that these promises are often overstated because data anonymization is notoriously hard. Where do we draw the line?

    Extensive monitoring deals will be hard for ISPs to resist since it often helps their bottom line to cooperate. Everyone seems to win, except for the users. Free speech is hindered when people know that they are being watched. Users already don’t have many options in choosing ISPs, and once monitoring becomes ubiquitous, there will be no good choices left.

  17. Harlan Yu Says:

    Fred: There aren’t that many protocols in widespread use. Sure, new ones get invented, but not that quickly that it would be impossible for a motivated ISP to keep up.

    You’re getting at an interesting and important point though. The idea of a census suggests complete knowledge about the tracks people are downloading or listening to. But it’s not technically that easy at the ISP level. They could try to notice metadata (say) in .torrent files or in id3 tags of mp3s, but this information is user-generated and often inaccurate or absent. They could try to identify the content itself (e.g., identifying specific blocks of a torrent download), but this is not trivial to do and assumes that the ISP has a huge dataset/repository of all songs, by artists large and small, to match the content against. No matter what they try, the census data will probably be incomplete.

  18. Richard Cant Says:

    How about allowing consumers to decide explicitly where a proportion (say 20%) of their license fee goes and then use that data (with a cut off of say 1000 donations minimum) to distribute the rest.

    The 20% plus cut off mechanism is needed to prevent people recycling their license fees through their friends.