Spotify Discovers the Marginal Value of Infringement in the Wrong Tail

[From The Huffington Post, January 18, 2016]

Spotify's music offering may have appeared to be the long tail, but now it looks like the wrong tail. After two class-action lawsuits for unlicensed songs and copyright infringement were filed against the company by songwriters in as many weeks, Spotify is doing back flips to blame the victim for its admitted failures to lawfully obtain mechanical licenses. But at the end of the day, the decision is governance not guidance, and Spotify's governance is solely in the hands of its corporate directors and officers. Those directors and officers must answer to shareholders for weighing the marginal value of millions of unlicensed songs over the infringement liability (not to mention an even worse press). 

Free Lunch or Free Will?

According to press reports, 10 percent to 25 percent of the songs on Spotify "are not properly licensed and/or not distributing royalty payments." Spotify also claims to have licensed approximately 30 million recordings (of 30 million songs, give or take for covers).

Based on these assumptions, that means there could be three million to 12 million songs that "are not properly licensed and/or not distributing royalty payments." This is not a few songs that fell through the cracks like new releases, a 1/16th of a song for a sample, the odd songwriter who cannot be found or who is non-responsive.

Millions of unlicensed songs doesn't look like an understandable accident, it looks more like an unacceptable policy. And that would be a policy that is exactly what the compulsory mechanical license was designed to prevent.

At the end of the day, the policy, i.e., the choice, to go forward without licenses, rests solely with Spotify. The well-financed company could have complied with the compulsory license -- enacted by the U.S. Congress for this exact situation -- but the songwriter class actions will likely seek to prove that Spotify chose not to do so. Whoever Spotify hired to undertake the mechanical process of mechanical licensing, millions of unlicensed songs strongly suggests that someone at Spotify decided to go forward without complying with the law. It appears that the thinking was that the upside value of having "all the world's music" was greater than the downside risk of getting caught. The marginal value of another few million songs was greater than actually complying with the law and paying songwriters.

This decision is what is called "business risk." Incredible as it may seem, this decision -- this willful decision -- to accept the business risk of using millions of unlicensed songs was apparently driven by a belief that in order to have an effective consumer offering, Spotify had to have tens of millions of tracks available to consumers. 

It's Not About a Database

You'll hear a lot of hand wringing (from many quarters including Spotify) about the lack of a central database that is constantly updated with complete rights information just so digital services can look up who to license from and who to pay. This database has never existed, but you know what does exist? The U.S. Copyright Office online searchable database of copyright registrations.

In David Lowery's case, the songs he complains of in his class action all were registered with the Copyright Office.  He published a list of them -- with registration numbers -- on his blog on October 20, 2015.  He also complained to the New York Attorney General about the problem in November with no response. 

Evidently, Spotify did not bother to look up Lowery's copyrights at the Copyright Office registration system as required by the compulsory license rules. That's behavior that is consistent with having millions of unlicensed songs. If the industry were to go to the trouble and expense of building this unicorn database, it would not have solved Spotify's problem with David Lowery because evidently nobody bothered to do the song research.

And this is the most elegant explanation of why Spotify is in the situation they are in -- they don't want to look because they don't want to pay the cost of complying with the law that provides them the great benefit of a compulsory license and they don't want to pay songwriter royalties. If Spotify really wants to pay songwriters, wouldn't they have done at least as much as the Copyright Office does with unknown writers -- publish a list online with all the information that they have (like artist name and song title).

Enter the Long Tail

This policy of using millions of unlicensed songs may well have been informed by the "long tail" theory and thought experiment posited by one Chris Anderson (in case you forgot him). You can read all about it in Anderson's counterintuitive utopian book The Long Tail: Why the Future of Business is Selling Less of Morewhich was based on a 2004 article in Wired.

I'd be very interested to know exactly where this consumer research is that shows the marginal value of an additional 12 million songs is so meaningful to a music service that it trumps the infringement exposure. I frankly have never seen it -- aside from Spotify's reliance on Anderson's version of the long tail.

Anderson goes down the wrong rabbit hole by relying on anecdotal observations of "Ben" an anonymized (or perhaps fictional) character (at pp. 3-5). "Ben" is a teenager from an affluent family in Silicon Valley who gets most of his music from "friends" and "Bit Torrent" (recall that Spotify's CEO was a developer of uTorrent, a key piece of the piracy picture acquired by Bit Torrent in 2006). So Anderson starts by analyzing a legal market with comparisons to the black market. That obviously wasn't going anywhere logical. Neither is any market of what the New York Times called "pixel-size niches".

Anderson's long-tail thought experiment has been criticized by a number of people such as Harvard Business School Professor Anita Elberse in the Harvard Business Review and most famously in the music business by Will Page, the former economist for PRS, the UK performing rights organization.

Any record company production manager could have chimed in -- and perhaps would have if it wasn't so obvious that it did not really bear much discussion. The corresponding transaction costs of a variety of functions including rendering royalty statements for minuscule unit sales were not worth keeping the title in the catalog. You know, kind of like sending a royalty statement for three streams. Preparing the statement may well cost more than the royalty even if the statement is itself digitally delivered. Not to mention taking the phone call from the angry songwriter who got a statement for $0.19.

Record companies are no strangers to the long tail -- that's often called classical and instrumental jazz. It is worth noting that record companies have for decades deleted titles that didn't sell enough to justify keeping the title in the company catalog. This is consistent with Professor Elberse's research demonstrating that "the tail increasingly consists of titles that rarely sell and that are produced by smaller-scale players." Professor Elberse assumed that there were no infringement costs associated with those "titles that rarely sell" thus exponentially increasing the cost of the tail, or as this particular tail is known in some circles, the wrong tail.

Is the Long Tail the Wrong Tail for Spotify?

Then-PRS economist Will Page reached a similar conclusion after analyzing PRS royalty payments in 2008. Those who have had about enough sanctimony from Spotify about how it is God's gift to fighting piracy will find this nugget of interest when wondering how much the marginal value of the last 12 million tracks that don't sell really is worth if they are all unlicensed:

Will Page, the former economist for PRS, found in his 2008 study that famously debunked Chris Anderson's absurd "long tail" theory, the "long tail" is pretty meaningless for music services:

[PRS] found that only 20% of tracks in our sample were 'active', that is to say they sold at least one copy, and hence, 80% of the tracks sold nothing at all. Moreover, approximately 80% of sales revenue came from around 3% of the active tracks. Factor in the dormant tail and you're looking at a 80/0.38% rule for all the inventory on the digital shelf.

Mr. Page joined Spotify in 2012, which was after Spotify's U.S. launch in 2011. If Spotify's policies led to its problems with songwriters, that is certainly not Mr. Page's fault, but Spotify could have just asked him about the cost/benefit analysis before continuing to take the business risk of failing to get compulsory licenses on what apparently is a gargantuan scale.

Was the marginal value of the long tail worth it for Spotify when compared to statutory damages? Commentators often mock statutory damages, especially for willful infringement, as being over the top. In the case of the compulsory mechanical license, you can look at this another way.

Congress did backflips to make the compulsory license easy to get. If a well funded company like Spotify (last valuation reportedly $8.4 billion) chooses to ignore Congress's efforts, then perhaps Congress wants to make sure that the marginal value of ignoring the compulsory license is always less than the statutory damages for choosing to do so.