Hacker News new | ask | show | jobs
by kwkelly 1637 days ago
Most serious online pollsters are using random sampling through panel providers and stratify based on known demographic statistics. Samples are re-weighted by raking (which reduces bias but increases variance) to better match population statistics.

This of course doesn’t mean all sources of bias are eliminated, but you also can’t eliminate all sources of bias in phone polling. Just like those taking polls on the internet, certain slices of the population that you can’t control for or don’t know about may be more or less willing to participate.