Are the polls producing a false five point Republican edge?

This interesting story comes from the McClatchy Newspapers:

Beware of polls that exclude cell phone-only voters

WASHINGTON—Watching the polls to figure out who’s up and who’s down this election season? Be careful. The poll may have a pro-Republican bias.

The ranks of Americans who use cell phones only have skyrocketed. Some public polls don’t survey them, however, and they miss a group of people who are more likely to vote Democratic, including the young, the poor, Hispanics and African-Americans.

The nonpartisan Pew Research Center recently found that polls that contact only those with land-line phones gave Republicans a 4 to 6 percentage point edge over Democrats compared with polls that included cell phones, in four out of five national polls this year.

In the most recent poll, a survey of likely voters reached via land lines gave Republicans a 12-point edge, 53-41 percent. Polls that also called voters who only use cell phones found the Republican edge was 7 points, 50-43 percent.

“Cell-only adults are demographically and politically different than those who live in land line households,” the Pew report said. “As a result, election polls that rely only on land line samples may be biased.”

The challenge in measuring public opinion has grown as more Americans rely on cell phones. In four and half years, the percentage of Americans 18 and older who rely on cell phones only has skyrocketed from 9.6 percent to 22.9 percent, according to the National Center for Health Statistics.

Similarly, the tally of households with cell phones only has risen from 10.5 percent to 24.5 percent. Another 14.9 percent of homes have land lines but report receiving most or all calls via cell phones.

Who are these cell phone-only people? According to the National Center for Health Statistics:

* The young: 49 percent of those aged 25-29, and 38 percent of those aged 18-24.

* The less educated: 25 percent of those with high school educations or less, the most wireless group.

* Students: 29 percent of people who go to school.

* Hispanics: 30 percent, the most wireless of any race or ethnicity, with African-Americans at 25 percent and non-Hispanic whites at 21 percent.

* The poor: 36 percent.

* Renters, 43 percent.

Many public polls this election season do include calls to cell phone-only voters, including McClatchy-Marist polls. However, some don’t, including automated surveys, such as the Rasmussen poll, that use computer-generated dialing and interviews.

“All things being equal, cell phones need to be included,” said Lee Miringoff, the director of the Marist Institute for Public Opinion at Marist College, which conducts the McClatchy-Marist poll.

“You don’t want to systematically exclude a group that may have different attitudes. It’s a large enough group that you can’t assume that they mirror the rest of the population.”

Calls to Rasmussen for comment weren’t returned.

Calling cell phones for a poll is more challenging and expensive than calling land lines.

First, Miringoff said, Federal Election Commission regulations require that they be dialed manually. Second, a questioner has to ask whether the person is driving or using heavy equipment. A “yes” answer requires a follow-up call later. Third, a high percentage of people with cell phones are younger than 18 and unable to vote. Fourth, more people refuse to answer the calls because they must pay for the air time; often pollsters compensate them, further driving up costs.

Overall, the “yield” of usable results from cell phone calls is about one-third of the calls to land line phones, Miringoff said. He cautioned, however, that the pro-Republican bias can be slight or overstated.

In Wisconsin, for example, Democratic Sen. Russ Feingold’s campaign last week disputed the McClatchy-Marist poll and others that found Feingold trailing his Republican opponent. One complaint: Polls were missing cell phone users and thus likely Feingold supporters.

However, Marist included cell phones in its sample. While Feingold trailed Republican Ron Johnson by 46-51 percent among land-line phone users, he trailed by a larger margin, 40-56 percent, among cell phone users.

“The cell phone users are not helping him at all. In fact, he’s trailing among them,” Miringoff said. “The problem for Feingold is not in the polling, it’s in the voter preferences. We’re seeing similar things in Colorado as well. The whole enthusiasm thing is more salient than the mode of communication.”

[end of article]

- end of initial entry -

Alexis Zarkov writes:

Nate Silver from fivethirtyeight has written on the cell-phone bias question here. Nate specifically treats the Pew Research study. As Nate points out most polling outfits try to correct for such effects by demographic weighting. For example it’s well known in the polling business that when you call a residence phone you are more likely to get a woman than a man. For some reason men are shy compared to women for taking phone calls. I know this myself because I have put in time doing phone bank work for a candidate. Most of the time a woman picks up. This is what’s called a coverage problem in polling. This means your sampling frame won’t get to everyone, meaning that not every member of the target population has an equal chance of getting polled. So the raw counts are adjusted to give more weight to men. The pollsters also do weighting by age, which might take care of the cell-phone problem.

The 2008 election seems not to have had a cell phone problem. Look for more of this type of story as the Democrats panic.

Posted by Lawrence Auster at October 19, 2010 08:28 AM | Send

Email entry

Email this entry to:

Your email address:

Message (optional):