Believe me, I'm well aware of the ways in which statistics can be misleading. (I'm taking an advanced research design class right now, and we're reviewing selection and sampling design flaws.)
The Harvard study results were more clear than I made them sound in my original post. See the details here:
Oberholzer-Gee and Strumpf's study used statistical models to compare downloads of songs from 680 popular albums in a variety of styles with the actual sales of those albums, as reported to Nielsen SoundScan, the industry's leading sales tracker. The researchers statistically analyzed whether the sale of an album declined more strongly in relation to the frequency and volume of downloaded songs from that album. In one week, for instance, they saw a spike in downloads of songs from the soundtrack to the film "Eight Mile." They would chart sales data for that CD in the following weeks to see if the download activity caused sales to decline.
It didn't, they were stunned to find. "This is where we cannot document any relationship between file sharing and subsequent sales," says Oberholzer-Gee, calling the effect "statistically indistinguishable from zero."
In fact, the study found that for the most popular albums - the top 25 percent that had more than 600,000 sales - file sharing actually boosts sales. For every 150 songs downloaded, the study showed, sales jumped by one CD.
I would be quite interested to know whether this 2004 study gets cited by the defense.