By Tricia Drevets
Hillary Clinton to win in a landslide.
Donald Trump narrows the election to voting in key states.
Depending on what news outlet you use, you can get a widely disparate view of the upcoming U.S. presidential election.
Why? News organizations rely on public opinion pills, and polls can be a faulty means for predicting voting outcomes.
Opinion polls, often just called polls, are designed to reflect the thoughts and likely voting behaviors of a certain cross-section of the population. Pre-election polls strive to reach a large random sample of voters, ask them a series of questions and then tally the results. Reputable polling organizations weigh in demographic factors in presenting the results.
Although there have been some major gaffes in the past – think the famous photo of President-elect Harry Truman holding the “Dewy Defeats Truman” headline – political polling held in the pre-digital age was successful for the most part. Throughout the years, polls have helped voters to know their opinions matter, and they have helped candidates solidify their messages.
However, the old means of polling is not working in 2016. In fact, the results of most of this fall’s polls reflect only one thing for certain – this is a divisive campaign. But we already knew that.
Opinion polls date back to 1824 and the presidential contest between Andrew Jackson and John Quincy Adams. That year, the Aru Pennsylvanian newspaper conducted a straw poll that showed Jackson led Adams by 335 to 169.
When Jackson went on to win the popular vote in Pennsylvania and then the entire country, other publications began to use polling methods as well. The Literary Digest, then a national publication, conducted the first national voters’ opinion poll in 1916. The magazine’s poll went on to correctly forecast presidential victors for the next four elections.
In 1936, however, the magazine blew it when it predicted that Alf Landon would triumph over Franklin Delano Roosevelt. Hindsight revealed that the Digest’s 2.3 million pollees had a decided Republican bent. When Gallup, still today one of the nation’s main polling organizations, took a poll of a more demographically diverse group of voters, it predicted Roosevelt's landslide victory.
Gallup is not immune to polling problems either, however. It was behind the mail-in poll that predicted Dewey would beat Truman in a landslide in 1948. Gallup’s error? Again, hindsight is 20/20, and the organization apparently stopped polling too soon on Election Day.
Over the years, polling moved from door-to-door-questioning, to mail-in postcards, to the use of landline phones.
Today, about 50 percent of American households do not have landlines, and even the ones that do have them, rely mostly on their cellphones and the Internet for their communication.
Since cellphones allow people to screen their calls, many American simply do not answer calls from pollsters. At Pew Research Center, the response rate to a telephone survey was 36 percent in 1997 and is only 8 percent today.
As a result, online polls are replacing phone polls. First of all, they are less expensive for the polling organizations, since they do not need to pay people to make the thousands of calls needed for a large poll. And secondly, they can go right into a recipient’s inbox when they want them to be there.
Internet polls do have some challenges, however. Mainly, since people easily can opt out of online polling, it is difficult to get a random sampling of voters. Some organizations are finding ways around this problem by collecting data from the same large randomly recruited group of voters for each survey they conduct.
The American Panel Survey (TAPS) uses this methodology. A TAPS panel at Washington University consists of 2,000 people, who are paid $10 per month to participate in online surveys. TAPS assembled the group through a random direct mail sampling of U.S. residents.
TAPS even reaches out to the 15 percent of its polling group who do not have Internet access by providing them with the tools they need for web access.
About 75 percent of the TAPS panel members participate in the polling each month. TAPS estimates that about 10 percent of the group drops out each year, so replacement members are added using the original sampling method.
Other polling organizations are using similar procedures for online polling. The Pew Research Center’s American Trends Panel and the RAND Corporation's American Life Panel are two examples. An interesting aspect of these targeted polls is that they can track how individuals’ opinions change over time, offering social scientists important “longitudinal” data.
Researchers are also experimenting with texting polls, especially for situations when a quick response is desired, such as “Who do you think ‘won’ the debate this evening?’”
Smartphone technology also opens up the possibility of directly sending voters campaign videos and ads to gauge their responses. Results from these kinds of polls may change the way candidates campaign and the messages they wish to convey.
Just how far will all this go? It is hard to say. Traditional telephone polling is still very much in use for the 2016 election, and many survey organizations are reluctant to abandon it completely. For one thing, online polls often don’t offer the immediacy of telephone polls.
TAPS participants usually have an entire month to answer polling questions, although about half of them respond within 48 hours. A month is an eternity when it comes to a volatile political campaign such as the one the U.S. has been experiencing this year.
“For elections -- an outcome for which pollsters can be held immediately accountable -- [online polling] isn’t performing any better than, and arguably worse than, standard telephone calls," Janine Parry, director of the Arkansas Poll said in a recent interview with Governing Magazine. "Phone surveys are still doing pretty well overall.”