How do we know if the music industry is doing well?

To re-iterate my extremist line against even [Mark] Cuban, however, aggregate sales aren't really all that relevant. If the music is getting made, and the music is getting listened to, well then, that's a healthy IP environment. The creation and consumption of new works is the end, sales are merely a means to that end.

Matt Yglesias continues:

I don't believe I've even heard anybody try to argue that fewer new songs are being written or recorded or that people are listening to less music than they used to. It's the availability of new music for consumption that we're supposed to be protecting here. P2P, through both authorized and unauthorized uses, obviously leads to an uptick in the number of people who listen to any given song. You would need a pretty huge decrease in the quantity of new music being recorded (and, as I say, no sign there is such a decrease at all) in order to make the case that the progress of musical arts was being seriously impeded here. P2P probably is a problem for the major record companies, both because infringement will reduce their sales potential, and also because it will make it easier for public domain and independent works to be distributed and publicized. This, however, simply isn't something IP law is supposed to prevent. The health of music-production as an endeavor is not at all the same thing as the financial status of the RIAA's membership. As I say, if there's evidence that thanks to copyright infringement kids aren't forming bands anymore, artists are quitting the business in droves to go to law school, or clubs are finding that nobody wants to go on tour anymore then that would be interesting and relevant, but I'm not familiar with any such evidence.

He truly ought to be an economist, albeit with more Milton Friedman driven into his blood (although here is his George Stigler impersonation). 

I'll add two points of my own.  First, file-sharing is in the long run a greater danger for capital-intensive movies; I worry less about music.  Second, most new CD releases fail to either earn money or attract real attention, with or without file-sharing.  Furthermore most consumers listen to only a subset of what they buy.  A first best probably involves lower music sales, not higher sales.  Let's say you sample a CD on iTunes (or illegally for that matter), don't like it, and then don't buy it.  The music company stops funding that kind of music.  Why is that market failure?  Admittedly we may be cutting into a cross-subsidy of some winners; the fools and their mistaken purchases help cover the fixed costs of the music company.  Still it is probably better to move closer to a better informed optimum, which again means less music not more.

Addendum: In addition to the excellent Mark Cuban link at the very top, Matt offers more data.