Wednesday, October 20, 2010

Cell Phones and Polling

is a lot of misinformation about cell phones and polling out there and the election season is only going to bring on more polls, so I thought it would be a good time to review some recent work on cell phones and polling.

Many people believe that pollsters cannot include cell phones in their polls. That’s simply not true. Many pollsters (and almost all of the good ones) include cell phones in their sample. Pollsters usually draw samples from what is called “random digit dialing.” Let’s take Sylva for example—many Sylva phone numbers begin with 586. A pollster then can call numbers at random beginning with 586. Cell phone numbers around here often begin with 399, 506, 507, or 508. There’s no reason a good poll can’t do the same thing with cell numbers that they do with land-line numbers—begin with a known prefix and then generate random numbers from there.

The problem with cell phones comes from people like my friend Eric. Eric lives in Sylva, but moved here from Tennessee. When he moved, he kept his Tennessee area code. A pollster who is trying to do a poll of Jackson County residents would not be able to include Eric. Similarly a person who leaves Jackson County may keep her phone number with a 399 prefix, and she might get a call for a survey meant for Jackson County residents.

So—how often does this happen? The Pew Center for the People and the Press recently released a report on this and found that 94% of cell phone users have a number that falls within the same Census region in which they live. 90% have cell phone numbers that match their current state of residence and only 60% have cell numbers that fall within their current county of residence. In other words, the smaller the geographic area you’re polling in, the bigger the problem you’ve got.

So what does this mean for people who want to make sense of polling data? First, make sure that the poll you’re reading about includes cell phones and not just land-lines. Polls that include only land-lines probably under-represent young people and Democrats. Second, the smaller the geographic unit, the more suspect you should be of the poll. As someone who’s conducted a few polls of local areas, I don’t mean to suggest that polls about small geographic areas are bad, but that a little more caution is probably warranted.

I’ll probably follow up with some more on polling soon.

No comments:

Post a Comment