It wasn't long ago that pollsters were fretting about a looming crisis in politics. It was called the cellphone, and its growing popularity threatened to skew surveys and ruin elections everywhere. Imagine not getting an accurate reading of a race because surveyors had only been able to reach wealthy, older people on landlines. At best, you'd risk giving voters the wrong impression of which candidate was ahead. At worst, you might unintentionally change the outcome of the race as campaigns scrambled to misspend their resources based on faulty figures.
Cellphones still pose a challenge for polling in some smaller races. But for the most part, the disaster that was supposed to be never took place, according to Scott Keeter, the Pew Research Center's director of survey research.
"Employing cellphones in the sample is more expensive," said Keeter. "We've been looking at that over the years and it's a problem that gets worse the more local your geography is. But for national polling — cellphones are not presenting a problem for us other than that they're more costly to call."
That's a break from even a few years ago, when Keeter told NPR that the exploding rate of cellphone adoption was "officially" skewing polls. Democratic pollsters complained as recently 2012 that President Obama's standings might be higher if polls did a better job representing cell phone users, which tend to lean more liberal.
To understand why cellphones were so problematic then — and why that may no longer be the case — we have to talk about an obscure rule that many feared would screw around with survey results.
The measure was designed to protect cellphone users from "robocalls," those phone calls where you pick up and hear an automated telemarketer pitching you something you don't want. For the ordinary American, getting these types of calls at home was bad enough without also getting them on their wireless phones. So Congress passed a law that, among other things, banned robocalling to cellphones. It also prevented the use of equipment that dials phone numbers automatically, one after another.
These well-intentioned restrictions had the unintended effect of tying pollsters' hands. Where previously they might have used robocalls to get a sense for how a race was going, they now had to use human dialers to do personal interviews lasting 15 minutes or more.
That extra human touch was expensive — in the beginning, interviewing a cellphone user cost roughly twice as much as a landline interview, said Keeter. But pollsters have since learned how to do things a little more efficiently, to the point where a cellphone interview now only costs 25 percent to 50 percent more. What helped drive prices down? For one thing, figuring out what times of day people on cell phones were most likely to pick up and talk. For another, understanding how to weed out cell phone users who are under 18 (because, obviously, underage users can't vote and aren't relevant to the polls). And finally, pollsters have picked up new techniques for getting cell phone users to open up.
Incorporating cellphone results also calls for slight changes in methodology. Cellphone users that also have landlines have a higher chance of being asked to participate in a poll compared to someone who has only a landline or only a cellphone, so researchers need to adjust their math to account for that. But Keeter says "we've pretty well licked" that statistical challenge. These days, 60 percent of Pew's poll respondents are cellphone users, and 40 percent are surveyed by landline.
And in some ways, having to conduct cellphone interviews by hand — so to speak — yields better information, particularly for outfits like Pew that are trying to explain an issue or election as opposed to predicting its outcome. Unsurprisingly, people respond differently to a human than they do to a machine, and because of it they're often willing to start talking in ways that you don't get with a robocall.
"You actually have a better chance of getting somebody on a cellphone to talk to you than you would if they were being bombarded by robocalls," said Keeter. "We're paying more — but maybe we're getting a little more."