Did you vote in the last election?
Of course you did.
At least that’s the answer most people give in a poll, whether they voted or not. It’s called the halo effect, which means that we want to be perceived as doing the right thing.
“You want to give good answers,” said Kerri Milita, who studies the science of polling. “You want to give the answers you think the person wants you to give. Basically, you lie.”
Milita, an assistant professor in the Department of Politics and Government, is familiar with the challenges of determining voter preference, but believes there are credible polling organizations, such as Pew Research and Gallup.
For a poll to be considered scientific, it needs to include a random sampling of 1,000 adults from across the country. That’s far different from a Facebook poll, or one conducted on the FOX News or CNN website where you can vote online or by phone. Milita said these informal polls are fun to look at but are the equivalent of a political horoscope.
There are polling websites that are valid. She rated election analyst Nate Silver’s statistics-driven website FiveThirtyEight.com as good. He accurately called all 50 states in the 2012 general election. She also favors RealClearPolitics.com, an often-cited source for U.S. public policy issues.
Polls are conducted by phone, email, and in person. The way questions are phrased, the order of questions, even the gender and race of the pollster can influence the outcome. And they’re only talking to people who are willing to talk to a pollster.
“Certain types of people will hang up. Certain types of people will not respond, and that biases the polls a bit,” she said. “Polling is a very imprecise science. There are so many nuances. That’s why we can’t rely on polls.”
Word choice can influence responses. She gave the example of using “very” versus “extreme” to describe being conservative or liberal. Because of an aversion to the word “extreme,” respondents will rank themselves closer to the middle. “How you word your question is as important as who you select to answer it,” she said.
As cell phones replace landlines, phone polls have become more difficult to conduct. Random digit dialing will find cell numbers, unless the person is on a do-not-call list. Online polling with random email addresses captures only those with Internet access.
National polls that attempt to predict the outcome of the presidential race are misleading because presidents aren’t elected by popular vote, Milita added. “They’re not going to battle nationwide, they’re going to battle state to state. We’d need 50 head-to-head polls to see how it’s going to come out.”
She also pointed out the difference between a poll that’s “reliable” versus “valid.” A poll can give the same answers over and over, which makes it reliable, but it may not be valid. Every adult has to have an equal chance of being sampled for it to be valid. Many polls reported in the media don’t meet that criterion, she said.
“I can put a poll on my Facebook page and it’ll give me the same answer. So it’s reliable, but that doesn’t mean it’s valid.”
Polls based on registered voters or likely voters carry a little more weight. And Milita likes exit polls, which tend to be accurate because people are usually honest about whom they voted for. Although she has never done exit polling, each semester she and Carl Palmer, an assistant professor in the Department of Politics and Government, take the political pulse of about 2,000 college students at eight universities across the country.
Drawing on her experience, she has a prediction for November. “This is going to be a tight one. Either way, it will be interesting.”