Calling the Presidential Election: Reforms to Polling Processes

By Eric Rabe, Fels Sr. Advisor

By 11:16 election night, President Obama called it. He happily tweeted “Four more years” and sent along a celebratory photo. Yet it was another two hours before Mitt Romney conceded and the end-of-the-campaign speeches were finally over. Romney’s polls were showing Ohio and Florida, two critical states, still could go his way.

In fact, the final results from Florida came in only on Saturday, four days after the polls closed. The state went for Obama by less than one percent of the vote.

“The numbers in Florida show this was winnable,” Romney adviser Brett Doster said in a statement to the Miami Herald. “We thought based on our polling and range of organization that we had done what we needed to win. Obviously, we didn’t….”

Nationwide, the final popular vote was indeed close,, 50.6% to 47.8%, but the electoral vote went to Obama in a landslide, 332 to 206.

Predicting that result proved difficult for those using traditional methods including some GOP-leaning pollsters. The closest polls got it right by relying on techniques that included Internet polling and sampled Americans, often younger, who have only a mobile phone.

Scientific polls have tracked elections for nearly two hundred years. The local straw poll in 1824, conducted by the Harrisburg Pennsylvanian, projected the result: Andrew Jackson over John Quincy Adams.

Having correctly predicted five presidential elections in a row, the Literary Digest poll famously flamed out in 1936 calling the election for Alf Landon of Kansas who wound up winning only Vermont and Maine.

Various opinions of what happened suggest that the self-selected sample was to blame. Furthermore, the sample was drawn from the magazine’s own readers, automobile owners and telephone subscribers, all possibly more affluent and more Republican than the overall electorate. That’s a notion challenged by some academic research, but in any case the poll’s inaccuracy led to serious study of scientific polling and the rise of the George Gallup organization.

During the later half of the last century, many accurate polls relied on telephone surveys. With more than 90% of Americans subscribing to phones at home, and phone numbers readily available from public listings, poll-takers had an effective base from which to sample the electorate. Today, with
traditional phone use in decline, polls that rely on calling voters at home may reach a sample that is older and whiter than the electorate at large.

In a convincing article, Nate Silver of the New
York Times FiveThirtyEight Blog argues that errors in polling during the 2012 election’s final weeks resulted from just this problem. His analysis shows that polls relying heavily on landline calls — like the Gallop poll — tended to show “a more Republican-leaning electorate than actually turned out” to vote. Silver, of course, accurately predicted the 2012 electoral outcome in each state.

Hopes of a GOP win may have been based on news media enthusiasm for a close election. They may have been based on Romney’s debate success and increasing popularity in the final weeks of the campaign. Or, they may have been the result of skewed poll results. Or, a combination of all these.

One thing seems clear. Because we Americans are changing the way we communicate with one another, pollsters will be reevaluating how they select and sample to accurately reflect the electorate. It won’t be easy. Polling organizations will have to reach voters who are young, mobile, concerned about their privacy and unwilling to make themselves easy to find.