Ed Felten on Freedom to Tinker hypothesizes a melding of several studies on file-sharing, creating A Grand Unified Theory of Filesharing. Copyfight noted the study here: Felten's Grand Unified Theory of File-Sharing. Felten divides the filesharing world into younger (15-24 years old) Free-riders (who fileshare and don't later purchase music) and older (25+ years old) Samplers (who fileshare to sample, but later purchase music).
However, while Felten's generational distinction is an important one, I'm not sure his theory fully explains what is going on. The main problem I see is that Eric Boorstin's thesis (Music Sales in the Age of File Sharing), which found that internet access correlates with increased music purchases for older people but decreased music purchases by younger people, isn't really about file sharing per se. The disconnect here is that there is no data for the correlation between filesharing and internet access.
In the "caveats" of his results section, Boorstin claims that:
However, it is highly unlikely that none of the measured effect of Internet access on CD sales is due to file sharing. There are many ways in which Internet access might affect CD sales, but the online activity that is by far the most directly related to the music industry is file sharing. On the negative side, MP3s are the nearest substitutes for CD purchases. On the positive side, the best way to learn about new music is to download that music. My results strongly correspond to the theoretical arguments and available survey information about the different file sharing effects of different demographic segments. Thus, file sharing is the likely explanation for my results.
Filesharing is one likely explanation, but not an exclusive explanation.
For example, older people use filesharing software much less than younger people. "File sharing is heavily skewed to youth. While a majority of Americans under eighteen have downloaded and half of those are heavy users, only a fifth of those aged 35-44 have downloaded files (Edison Media Research, 2003)." (The Effect of File Sharing on Record Sales: An Empirical Analysis [PDF] by Oberholzer and Strumpf). If this is so, then it may be that while internet access increases music sales in general, use of filesharing software decreases music sales across all demographics. Since older people have more access to the internet, but don't fileshare nearly as much, the overall result shows older people buying more music. In contrast, as younger people are more likely to use filesharing software, the positive effect on music sales of internet access is outweighed by the negative effect of using filesharing software.
Of course, this wouldn't be the only factor to distinguish between younger and older internet users (for example, the fact that older people tend to have more disposable income would likely have an impact), but it could be a significant one.
Boorstin found that increased internet access correlated with increased CD sales, not that increased filesharing use correlated with increased CD sales. I know that the internet (absent filesharing) increased my CD purchases, mostly due to convenience. Amazon is a great way to immediately buy CDs, either as an impulse purchase ("Oh, yeah, I should buy that album" - click, buy) or to sample music before purchase without downloading (many CD sale websites allow users to sample 30 secs. of streaming music or so). I've also purchased CDs from bands whose music I downloaded from the band's website (which doesn't really count as filesharing, I think). Obviously, my experience is anecdotal but, as Boorstin's research shows, internet purchases have increased substantially despite an overall decrease in music purchases.
It would also be interesting if Boorstin would revisit his numbers taking into account not simple internet access, but broadband access. I'm not sure what the results would be, but they might be illuminating if, for example, broadband access was correlated with increased or decreased music sales in comparison to dial-up access. Basically, to really understand what is going on, more statistical information is needed regarding filesharing.
Of course, as Felten notes, even if we knew what is going on today, it wouldn't necessarily be clear where the future leads.