Once upon a time, people had phones hard-wired in their houses. These phones were called “landlines,” and they relied on wires strung throughout their communities to operate. It was an effective technology, but one that necessarily meant that your ability to place and receive a telephone call was limited to a particular vicinity. Over time, inventors came up with a thing called a “cordless phone,” which extended that range slightly, but not much.
For pollsters, this was useful. There was a phone number that was connected to a house where you knew certain people lived, and so you could call that number and have a good sense that you were talking to Joe Smith, registered Republican.
Then capitalism got in the way. The invention of the cellphone and the rapid adoption thereof meant that people were no longer tied to a particular physical location when making calls. About a decade ago, as more Americans began relying solely on cellphones, pollsters began incorporating those numbers into their pool of contacts. This was tricky for several reasons, including that federal law mandates that cell numbers be hand-dialed. As The Washington Post’s Scott Clement explained a few years ago, this means that it can cost twice as much to call cell numbers. In the already-tight economics of polling, that’s a problem.
A decade ago, though, only about 1 in 8 adults lived in wireless-only households, according to data from the National Center for Health Statistics. As of the second half of 2016, though, slightly more than half of American adults fit that description for the first time. The most recent figure, for the first half of 2017, established that 52.5 percent of adults live in wireless-only households.
Generally speaking, pollsters are ill advised to ignore cellphone users, if only because they’d be missing half of the country. But there’s another reason that pollsters need to include cell users: The demographics of those with and without access to landlines is stark.
Consider race and ethnicity. Nearly two-thirds of Hispanic adults in the United States live in households that are wireless-only. More than half of black adults and Asian adults do, as well. But fewer than half of white Americans do.
Including only landlines in polling — which, we will note, is not common practice at this point — means you’re much less likely to reach Hispanic voters.
Or younger ones. Nearly three-quarters of adults ages 30 to 34 live in wireless-only households. Those younger than 25 are less likely to — probably because some chunk of that group lives in a household with someone age 45 or older (that is, a parent). Among those 65 and up, fewer than a quarter live in a wireless-only household.
There’s a trend undergirding this. Those under the federal poverty level are much more likely to live in wireless-only households than those earning at least twice that level.
Income, age and race all correlate to another factor: homeownership. If you own a house, you’re much more likely to have a landline in that house, both because older Americans still have landlines/are more likely to own houses and because people who rent are less likely to have a landline installed.
Owning a home tends to correlate to another characteristic: being likely to vote. If you have lived in the same place for a decade, you’re probably pretty familiar with where you need to go to cast your ballot — and it has been a long time since you’ve had to register at your new address.
The idea that the mode of reaching someone in a poll affects the results isn’t just a theory. In December 2015, we looked at research that showed spreads of more than 20 points in presidential and congressional polling depending on whether the respondent was reached on a cellphone or a landline.
That discrepancy, incidentally, is precisely why the National Center for Health Statistics collects this data. In an interview in 2015, an associate director for the agency explained to NPR that there were significant differences between the two populations.
“People who are wireless-only are more likely to smoke, they’re more likely to binge drink, they’re more likely to be uninsured,” he said. “In effect, they are more likely to engage in risky behaviors.”
Polling suggests that they also tend to vote differently.