May 27, 2018

Compulsory Licenses, and the Measurement Problem

At this week’s Future of Music conference, Terry Fisher of Harvard suggested yet another variation of compulsory licensing for online music. The basic idea is to slap a tax on computers, or on Net access, or on something else you need to get music online. Then the taxpayers can listen to all of the online music they want.

Seth Finkelstein points out, correctly, that the devil is in the details. These schemes don’t seem nearly so elegant when you flesh them out fully. Still, as I have written before, the compulsory licensing meme is so persistent that it deserves serious analysis.

Nearly all compulsory license plans split up the revenue among different copyrighted works, according to the size of the audience for each work. The audience sizes are estimated by random sampling. Fisher’s plan, for example, would embed a watermark in each work, and then use the watermarks to tabulate usage.

Finkelstein objects that this kind of scheme would require a ban on non-watermark-compliant music players, to make sure that usage is properly counted. I’m not sure this is right. As Neil Netanel has pointed out, most people are likely to want their usage counted, so that the artists they listen to get a bigger share of the pie. Because of this, most users won’t want noncompliant players. And if the watermark doesn’t try to be nonremovable, the engineering cost of reading it will be low. So it seems that most players will participate in the counting process, even if it isn’t required.

There are at least two problems relating to this kind of measurement, though. First, users will have an incentive to over-report or mis-report what they listen to. Sure, I would like to see money go to my favorite artists. But I would like even more to see it go to my brother, so I have an incentive to claim that I listened fifty times to my brother’s off-key rendition of “Feelings.” Worse yet, I have an incentive to erase the watermark from a Britney Spears song, replace it with the Peter-Felten-sings-Feelings watermark, and then distribute the Britney song like crazy, so that Britney-lovers boost my brother’s income.

In some ways, music usage data would resemble TV ratings, which also try to estimate home media-usage habits of ordinary people.

Back when the Nielsen TV rating service asked homes to keep TV-watching diaries on paper, there were persistent reports of people writing down what they wished they had watched instead of what they did watch. They might claim to have watched their favorite show every week, even when they didn’t; or they might omit lowbrow or unfashionable shows. They might claim to have watched a show if their friends liked it and it was in danger of being cancelled. (You might argue that these biases improved the overall quality of TV. That may be true, but there is no doubt that they caused the ratings to reflect something other than the actual audience size of the shows.)

I realize that some existing compulsory license regimes rely on sampling. But it’s one thing to sample what is played in a public setting, and another thing entirely to sample what happens in people’s homes. The measurement problem for compulsory net-music licenses is not insurmountable, but I think it needs more thought than it has gotten so far.