From Scalzi:
Also, these are self-reported numbers from about 5,000 North American authors, which is a) only a small slice of those writing books, either full-time or part-time, b) represents those who knew about the survey and were motivated to answer it.
Point a is not a real problem, survey sampling works pretty well with small sample sizes (and this doesn’t seem that small) if you don’t accidentally or intentionally choose a biased sample. So point b and self-reporting biases are much more valid complaints.
