Caller ID and Polls

Powerline links to some interesting info on polling methodology, which examines whether current polls overstate the amount of democrats around the country (as they did in the Bachmann Wetterling race).

However, I think a more interesting point is the impact of technology on response rates. I have seen some sources report that response rates to polls have dropped from 30% to 20% since the 1980s. What effect, however, does this have on poll accuracy. This article states that it has no effect.

I am not so sure. In my own unscientific survey on this subject I have found a couple of common themes. These are that married people with kids tend to take advantage of Caller ID more than single people and older people. As for myself, I would never fall into a poll sample because I screen every call. If it is a number I do not recognize, then I don’t answer the phone. In my unscientific survey, most other people I spoke with in my demographic (married with kids) behave the same way – the screen all their calls and thus would never be included in a poll In contrast, older people tend to not have caller ID or ignore it and a lot of single people I talked to also said they answered the phone without paying attention to Caller ID. Are older and single people then oversampled when compared to married people with kids?

Caller ID has changed the landscape of opinion polling in the last ten years mostly because 1) it is very inexpensive (my digital phone service includes it for free) and 2) almost every phone you can buy now has a caller ID window on the handset so seeing the ID is much more convenient and cheap than it was 5 years ago. It is possible that polls over sample certain groups because of Caller ID. Further, it is also very likely that these oversamples include more democratic voters, which skews the polls.