You make a very articulate point, but let me disagree on this: It isn't true that a poll like this is "useless for modeling reality". It tells us the distribution of voluntarily given responses about the geographical distribution of registered users of HN who happened to see the poll and happened to want to answer.
How does this correlate to the absolute geographical distribution of readers? Or of registered, but maybe inactive, users? That is difficult to guess. But the information from the poll is, for me, already very interesting.
By the way, I'll add that even with polls you can always only measure voluntary responses, even if the sample is randomly extracted, and there is no way around this.
How does this correlate to the absolute geographical distribution of readers? Or of registered, but maybe inactive, users? That is difficult to guess. But the information from the poll is, for me, already very interesting.
By the way, I'll add that even with polls you can always only measure voluntary responses, even if the sample is randomly extracted, and there is no way around this.