its called selection bias... (Score 1)

The number one defacto problem with sample based studies is that we know for a fact that people who take part in surveys are not necessarily representative of everyone else. The magnitude of this problem is debated (see John Brehm's book The Phantom Respondents).

Read the social science classic Designing Social Inquiry: Scientific Inference in Qualitative Research by King, Keohane, and Verba.

Chapter 5 on "What to Avoid" explains how selection bias works, and why, for example, asking 40,000 people who agree to have their computers monitored by a commercial research marketing group is probably a heavily biased sample.

That being said, its interesting to note that the same research firm notes that while file sharing has an impact on record sales, the music industry is still to blame for declining sales.