Standard errors from nonrandom samples

Simon Jackman has some useful thoughts on the future of internet polling and why some of its critics don’t know what they’re talking about.

I’d also like to add that, as much as I rely on telephone sampling in my own research, I hate most of it in practice. Just yesterday, the phone rang and it was one of those robo-polls. I hung up, but I hated to even have to waste 5 seconds on the intrusion. I’m bothered by the asymmetry, that the survey wastes the time of the person who is being called while imposing a nearly zero cost on the pollster. Information pollution is what it is.

1 thought on “Standard errors from nonrandom samples

  1. More interesting than Jackman's thoughts is the fact that nobody commented in the first 24 hours.

    The dirty secret to polling is that the cooperation rate is way down from what it was when telephone surveys largely replaced personal interviewers. Jim Spaeth from the ARF estimated the response rate was about 14% a few years ago. With caller ID, cell phones and a general glut of pollsters asking for everyone's opinion, it's probably lower now.

    Measuring these biases and correcting for them is a nontrivial and rather inglorious job, and I don't know of a particularly good text for this, despite the fact that this is most of what I've done for years.

Comments are closed.