Stardate 20020424.1818

(On Screen): Jupiter Media Metrix reports that file-swappers buy more music, not less. It's a black eye for RIAA.

Not so fast. They didn't prove anything of the kind. Their claim may be true, but what this poll showed is that file swappers said that they were buying more music. Given the political situation, and given that file swappers are particularly attuned politically on this issue, they have a strong incentive to lie and claim that they were buying more even if they weren't, so as to head off relevant legislation.

Too many polls make the underlying assumption that the people they talk to actually tell the truth. Which is odd, because history is full of examples which prove otherwise. But on that the results are extremely variable: on a lot of kinds of polls, responses tend to be very honest. On others, the rate of lying can exceed half. So you can't always tell whether a poll really means anything. When they say "the error on the result is within plus or minus 3 percentage points" they're talking about sampling error as a function of the size of the group polled. That has nothing whatever to do with the rate of veracity, which can't easily be estimated.