Study: Cell phones tilt polls
This month's election results lent support to what many pollsters have long suspected: Polls that don't include cell phones favor Republicans.
And as more Americans come to rely on cell phones, the disparity is widening, according to a study released Monday by the Pew Research Center.
Pew’s surveys that included cell phones were dramatically closer to the actual midterm election results than they would have been without cell phones in the sample, according to the report.
A national Pew poll conducted days before the election found likely voters preferring a Republican to a Democrat for their representative in Congress by a 6-point margin when both land lines and cell phones were surveyed.
That margin closely matches the national vote for House candidates, according to the latest count: Republicans led by a 7-point margin.
But if the Pew pollsters had called only land lines, disregarding their cell phone sample, they would have found the GOP ahead by 12 points.
The new study bolsters the argument that polls that don't call cell phones will be systematically biased. Calling cell phones poses numerous challenges for pollsters — from the difficulty of determining where respondents live based on their cell phone numbers to the fact that machine-aided dialing of cell phones is illegal.
While previous Pew reports had found a gulf between its overall sample and its land-line-only sample, the election result shows which one was right. The winner, clearly, was the sample that included both.
The election result "is a very good opportunity to evaluate the effectiveness of your methods," Pew's director of survey research, Scott Keeter, said. "We don't have that many opportunities to have that kind of validation."
And while Pew looked for evidence of a gap only in its own polling data, an independent analysis found a similar gulf existed at polling firms using different methods during the 2010 cycle.
The six national polling firms that surveyed both cell phones and land lines in the week before the election put Republicans ahead by an average of 6 points, while the four that called only land lines showed a GOP advantage of 10 points, according to the analysis by Mark Blumenthal of The Huffington Post.
Pew has been tracking the issue of potential cell phone exclusion bias for several years, but it is only recently that the gap has widened into a statistically significant difference. What began as a slight bias in favor of Republicans has steadily increased into a marked GOP skew when cell phones are ignored.
Pew's late surveys in the 2008 presidential race put Barack Obama ahead of John McCain by 8.2 percentage points among all types of phone users but just 5.8 points among land line users only, for a pro-Republican tilt of 2.4 points in the land-line-only sample. (Barack Obama actually won the national popular vote by a margin of 7.3 points.)
In Pew's last three polls leading up to the 2010 elections, however, the tilt toward the GOP of the land line sample had more than doubled, to 5.1 points.
As cell phone use increases and land lines dwindle, the trend is clear, Keeter said, and likely to continue.
The Pew study divided those polled into three groups: those who have only cell phones; those who have only land lines; and the largest group, "duals," who have both. Interestingly, the study asserts that even people who have land lines aren't represented accurately in land-line-only surveys.
"What the report is suggesting is that among people who have both land lines and cell phones, if you reach them on their cell phone, you get a different kind of person," Keeter said.
Pollsters who call only land lines are obviously missing the approximately one-fourth of adults who don't have land lines. But they're also missing those people who have both but pick up only their cell phone, he said.
The "duals" polled on their cell phones tend to be younger and less white than "duals" polled on land lines, according to the study.
The rapid expansion of American cell phone use coincides with the rise of inexpensive automated or "robo" pollsters. One of the major reasons the robo-pollsters are regarded with suspicion by traditional live-interview polling firms is the fact that they don't call cell phones.
"The inefficiency of calling cell phones right now, just in a monetary sense, puts cell phone calling out of the reach of some organizations, or at least their clients," Keeter said. "We did this to demonstrate what the situation is, that including cell phones is important because the population that can't be reached by land lines is politically significant and relevant."
According to federal survey estimates, 25 percent of the U.S. adult population relies exclusively on cell phones, while just 11 percent of households have only land lines. "Cell only is now twice as numerous," Keeter said. "Not long ago, those numbers were virtually reversed."
Young people in particular are becoming more and more difficult to reach on land lines. Despite making multiple calls over four or five days, Pew's raw land line sample is just 7 percent people under 30, who represent 22 percent of adults. Hispanics, less wealthy Americans and less educated Americans also are growing harder to reach on land lines.
The bottom line, Keeter said, is that cell phones are an issue for the polling industry that is only going to grow in importance.
"This is a challenge that threatens polling's ability to explain the election, not just predict who's going to win," he said. "If you are missing almost completely a key demographic group because your methods can't get them, then you're crippled in your ability to tell the story of the election."
POLITICO
No comments:
Post a Comment