Do Polls Undercount Obama Supporters?

If they fail to keep up with changes in telephone usage, they might, says one pollster.

Image Source/ Getty Images
Image Source/ Getty Images

(The Root) — In the weeks before an election, much of the media coverage becomes a numbers game, obsessively chronicling which candidate is in the lead in which states and with which voter blocs, according to the polls. But one question that media and consumers of media rarely ask is, how do we know the polls we are reporting and consuming are accurate? Increasingly it appears that traditional polls may be missing wide swaths of specific voter demographics altogether, particularly younger voters of color.

Polling data play a significant role in elections, particularly in the Internet age. Today a polling firm can release new data online that can reach millions within an hour. The impact can be virtually instantaneous and far-reaching. For example, four years ago then-vice presidential candidate Sarah Palin had one of her most public clashes with John McCain’s campaign staffers after she publicly criticized the campaign’s decision to stop campaigning in the state of Michigan, thereby surrendering the potential swing state to then-Sen. Barack Obama’s campaign a month before Election Day.

The campaign’s reason? Polls had begun to show Obama with a lead, so the decision was made by the McCain team to start investing the resources being poured into Michigan into more viable states. (Recent polls indicate that history may repeat itself, with Mitt Romney now also trailing Obama in Michigan, the state in which the GOP presidential candidate was born.)

While advancements in technology may have helped increase the reach and influence of polling, so far technology has not been fully utilized to ensure the accuracy of polling, which still relies heavily on lists of landline telephone numbers to reach people. As of 2011, 83 percent of American adults owned a cellphone, and the numbers of Americans using one as their primary form of communication is growing rapidly. Between 2007 and 2010 the number of American homes without a telephone land line doubled, to 26.6 percent. Now a full third of U.S. households are cellphone-only.

Among young people the numbers are more pronounced. More than half of adults ages 25 to 29 now live in wireless-phone-only households, according to an FCC report. The Centers for Disease Control and Prevention’s National Center for Health Statistics referred to a growing number of young people as “cord-never-getters,” meaning they get a cellphone and that becomes their primary phone for life, so they never bother getting a landline even when they move out of Mom and Dad’s house.

Despite the explosive growth of cellphones as the primary form of communication for a number of Americans, political polls are still largely conducted via landlines. The reason is largely cost, according to David Paleologos, director of the Suffolk University Political Research Center, one of the most quoted political-polling institutions in the country. Polling firms purchase lists of phone numbers in order to reach respondents. Cellphone-number lists are significantly more expensive than lists of landline phone numbers — between double and triple the cost.

Additionally, cellphone polling is more labor-intensive because, for example, if you are trying to conduct a poll of likely voters in the Massachusetts U.S. Senate race between Elizabeth Warren and Scott Brown — something Suffolk recently did — cellphone-number lists will include a lot of young people who lived in Massachusetts for college but have moved on. This means pollsters have to keep calling cellphone lists until they reach enough people they can confirm live in the area relevant to the poll.

“The media wants the best of both worlds,” Paleologos said in an interview with The Root, referring to media outlets that commission polls during campaign season. “They want the best data, but they want to keep costs manageable.”