An NPR member station
Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations

A researcher explains why polls failed to predict a Trump victory

SCOTT SIMON, HOST:

Election Day is when Americans come together to decide the future of the country. The day after Election Day is when Americans wonder, how did the polls get it wrong again? Sunmin Kim is an assistant professor of sociology at Dartmouth College and one of the authors of a recent study on why public opinion polls so often fail. Professor Kim joins us from Hanover, New Hampshire. Thanks so much for being with us.

SUNMIN KIM: Thank you, Scott.

SIMON: The polls predicted a tight race. It wasn't. What do you think or who do you think pollsters missed?

KIM: So I don't think polls technically missed this outcome that we have seen, because polls always specify the possible range of the outcomes. Certainly, there are things that the polls fail to notice or we fail to notice within the polls. And those are the things that we have to sort of, like, focus on in assessing the instance like this one.

SIMON: Well, what factors were missing, do you think?

KIM: One thing that I'm focusing on is the increase of Trump supporters in the already-Republican districts, places like Florida and other parts of the South. Trump support have increased by a substantial margin, and the same goes for the place like California. There was a substantial increase of the Republican support in the California, especially in the inlands. But we don't have enough data to make assessments on what happened in the last four years in those communities.

SIMON: What are some of the practical limitations of conducting a poll?

KIM: They are increasingly doing a better job of adjusting for various kinds of biases and sampling. But there's no way to counter this propensity for the respondents to not take the call. So when you are exposed to the polling relentlessly, as you do in the election season, especially when you live in the swing state, people tend to sort of, like, hang up when you get a call from the pollsters.

And another thing that we have a hard time discussing is what scholars call the social desirability bias. People are reluctant to share things that they are ashamed to share. And in this case, you know, supporting the President Donald Trump configures perfectly into this social desirability bias. And those things are, you know, like, are crucial limitations when it comes to assessing the support for someone like President Donald Trump because if you look at the elections when he was on the ballot, 2016 and 2020, that coincides with the largest polling misses in the history of an enterprise.

SIMON: I'll bet not an hour of our programming goes by that doesn't cite one poll or another...

KIM: Yeah.

SIMON: ...So that we can say, well, according to the polls, Americans believe this. Americans believe that. Campaigns use polling information in many ways to direct themselves, what issues they're going to use to shape their message. Should they?

KIM: Well, they should because there are practically no better alternatives. But I think we should take polling with more caution, as scholars have long been advocating. I, for one, after witnessing the outcome of this election, is leaning towards a more in-depth reporting or the ethnographic studies of particular communities because when you do the national representative or statewide polling, we often neglect specific dynamics that occur in the underground in the communities, and we are left wondering with the numbers by subgroups. But they are not knowing what these subgroups went through in the last four years. And that's why I think, you know, this more engaged, in-depth journalistic reporting of the communities might be more useful in this new polling environment, where people are more hesitant to express their opinions. And it's hard to sort of, like, get their attention amid this already saturated information environment that we live in.

SIMON: What advice would you give people when they read a poll?

KIM: Well, take it with a grain of salt. Pay attention to the margin errors because the center point is not the precise estimation of what's happening out there. It's a center point of a range of estimations. It's convenient to, like, just conduct a statewide or nationwide polling and make an inference based on the data, but it should always be triangulated with the other forms of the data.

SIMON: Sunmin Kim is a professor at Dartmouth College. Thank you so much for speaking with us.

KIM: Thank you so much. Transcript provided by NPR, Copyright NPR.

NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.

Scott Simon is one of America's most admired writers and broadcasters. He is the host of Weekend Edition Saturday and is one of the hosts of NPR's morning news podcast Up First. He has reported from all fifty states, five continents, and ten wars, from El Salvador to Sarajevo to Afghanistan and Iraq. His books have chronicled character and characters, in war and peace, sports and art, tragedy and comedy.