Listen To This Story
The 2016 election of Donald Trump changed everything. Polls showed Trump wasn’t supposed to win. Hillary Clinton was. Nebraska Public Media’s Dale Johnson reached out to Kristen Olson, an expert in survey methodology in UN-L’s Bureau of Sociological Research, to talk about changes in the science of polling since 2016.
Dale Johnson: Why were polls wrong in 2016?
Kristen Olson: I served on the American Association for Public Opinion Research Task Force to evaluate the 2016 election polls. We had a variety of conclusions. So the first is that the national polls were pretty good. Actually, historically, they were pretty good. The big difference was that the popular vote and the electoral college went different directions. So in 2016 the popular vote was for Clinton and the Electoral College was for Trump. The state level polls were off, and that's what determines the Electoral College. They were off in ways that are complicated. State level polls overrepresented people who had higher levels of education. So more people with college degrees were in the polls than in the general population. Those people tended to vote for Clinton, and so since there were more of them, it looked like Clinton was going to be the winner in those polls. The state level polls didn't account for that statistically. After statistical adjustments to weights were done, those state level polls came closer to the actual vote counts. The other thing that happened in 2016 was a lot of people decided late. People were making up their mind who to vote for on the day of the election, the few days before the election, the polls don't reflect those late decisions. The polls reflect what people thought yesterday, but they don't know what people are going to do tomorrow.
Johnson: Polls only capture a snapshot in time.
Olson: Absolutely.
Johnson: So that begs the question, how long should a poll be trusted?
Olson: Understanding the time frame that something was conducted is really important. Some polls or surveys have particular references. So if you were to vote today, who would you vote for? So it is less about trusting the polls as a predictor of what's happening in the future, but more as understanding what happened during the time point under which the polls were conducted.
Johnson: Why should any attention at all be given to polls, and I say that with a smile, and the result is a smile on your face. Why should polls have any attention paid to them?
Olson: It's really important to understand what people think, feel and do. And it's really important to understand how people think, feel and do things across a wide swath of the public. Polls and surveys give us a way of hearing the voices of many on things that are really important to people. The horse race questions, who you're going to vote for, what you're thinking about particular initiatives are only part of the stories. Understanding why people think those things, what's factoring into people's choices, how people are making those decisions. Those are equally important, and are also part of many polls that are done. Those are really important to know, because you can't find that out on election day.
Johnson: There will be a lot of polls between now and Election Day. Your advice to listeners about the next poll they see, hear or read about?
Olson: My advice is to pay attention to who did the poll, why they did the poll, and what the questions were that were asked for the poll. I would also say that consumers of polls should realize that that is not the end result of the election, that there's a lot of time between whenever a poll is done or a survey is done and the election day, and so those that information is not the end story.
Johnson: We'll go back to a term we used earlier in the conversation, a snapshot in time. Listen to when that poll was conducted.
Olson: Exactly.
Johnson: Kristen, thank you for your time. Kristen Olson joining me, an expert in survey methodology in UNL’s bureau of sociological research, talking about polling today. I'm Dale Johnson on Nebraska Public Media.