Over the last decade, as mobile phones have become more prevalent and a quarter of American adults have shed their landlines, pollsters have debated how important it is to include cell phone users in their polls. Now, a Pew Research Center report finds that polls in the 2010 election cycle that contacted people on landlines only overestimated Republican support by as much as 6 percentage points. The reason: those who use cell phones only are on average younger and more likely to be minorities; both groups are traditionally Democratic voters.
Republican Pollster Adam Geller of the National Research Group does not believe that excluding cell phone-only voters poses as big a problem as the Pew report suggests. “There are a half a dozen reasons why a poll can be off and you really can’t ascribe it to a poll that does or does not dial a cell phone,” says Geller, adding that standard practices such as weighting for anticipated turnout can eliminate the bias inherent in a landline-only poll.
Few pollsters would dispute that cell-phone-only households tend to be younger and are more likely to be single and members of minority groups. However, some pollsters do question how likely people in these households are to vote—especially in a midterm election. “What we are probably describing by and large is a more presidential year voter,” says Geller. He does concede that for a presidential year, when turnout is generally higher and pollsters use less stringent likely voter screens, it makes sense to include cell phone-only households.
Mark Mellman, CEO of the Mellman Group, a Democratic pollster, counters that polling cell phones is imperative to achieving an accurate sample today and that polling firms that fail to include cell-phone-only households in their samples should be viewed with suspicion. “We were the only poll that was right about the Nevada Senate race,” he says. “There was a vast difference between cell phones and landlines, and if we had just polled landlines, we would have had the same inaccurate results.”
Mellman says that there are some regions of the country where including cell phones can make a crucial difference in poll results, and there are some regions where it is inconsequential. The problem there is that there is no way to know for sure yet which sort of region you are dealing with before you conduct a cell-phone poll. Mellman adds that the effort is always worth the expense. “If you do not poll cell phones you cannot be sure you are getting accurate results,” he states. “Period. End of story.”
The law requires that cell phone numbers be hand dialed by live callers, which means that polls including cell phones require more time spent in the field. (This is also why pollsters that use an automated system, such as Rasmussen Reports, do not include cell phones in their samples.) Going to this extra trouble in a non-presidential-election year, according to Geller, increases costs considerably without yielding dramatically different results. For his part, Mellman believes that the choice between polling cell phones and not is stark. “It is more expensive, it is more difficult, but if you care about getting it right, you will expend the effort,” he says. “People who care about getting it right will do it.”
The 2012 cycle will undoubtedly see many polls that include cell phones, but it is not yet clear whether the same will be true of polls in years between presidential contests.
Noah Rothman is the online editor at C&E. Email him at firstname.lastname@example.org