For the 53rd episode of the Journalism History podcast, host Ken Ward spoke to Wendy Melillo about how political polls have been used—and misused—and the election prediction mishaps in American journalism history that followed.
Wendy Melillo is an associate professor at American University, where she teaches in the School of Communication’s Journalism division. Her research is focused on how strategic communication influences society and the media. A former reporter for The Washington Post, Melillo earned a Pulitzer Prize nomination and an award from the White House Correspondents’ Association for her coverage of the United Way scandal. She spent nearly a decade covering political advertising and strategic communication planning for the business publication Adweek.
This episode is sponsored by Taylor & Francis, publisher of Journalism History.
Wendy Melillo (00:04):
So if we understand this in the context of the history repeating itself, we end up back to where we started our discussion because the disconnect between what polling was intended to reflect—the will of the American people—and how polling has come to be used in modern day America—to predict election outcomes—you know, continues.
Ken Ward (00:31):
Welcome to Journalism History, a podcast that rips out the pages of your history books to reexamine the stories you thought you knew and the ones you were never told.
Teri Finneman (00:40):
I’m Teri Finneman, and I research media coverage of women in politics.
Nick Hirshon (00:45):
And I’m Nick Hirshon, and I research the history of New York sports
Ken Ward (00:50):
And I’m Ken Ward, and I research the journalism history of the Great Plains and Rocky Mountains.
Ken Ward (00:55):
And together we’re professional media historians guiding you through our own drafts of history. Transcripts of the show are available online at journalism-history.org/podcast. This episode is sponsored by Taylor and Francis, the publisher of our academic journal, Journalism History.
Ken Ward (01:12):
The night of November 8, 2016—that’s Election Day 2016—was an interesting one, regardless of your political beliefs. Most of us went into that evening believing what the polls had predicted: a victory for the Democratic nominee. Yet as the evening progressed and the results filtered in, the accuracy of those polls came into question, and for good reason: despite the supposed clarity of the polls, the results showed a clear electoral college victory for the Republican nominee. The fallout for pollsters was swift. How could the polls have been so wrong? What happened?
But as explained by Wendy Melillo, associate professor of journalism at American University, polls have frequently misled the public and pundits alike when used to predict election outcomes. In fact, the way political polling in the United States is used today and for much of its history barely resembles what its creators intended. Wendy joins me to explore this history as we follow political polling from its inception to today. Wendy, welcome to the show.
Wendy Melillo (2:12):
Thank you so much for having me on.
Ken Ward (2:15):
Absolutely. So, first question here, when did political polling in the United States become widely practiced and how did the people who came up with polling in the first place envision it actually being used?
Wendy Melillo (02:29):
Okay. So let’s first look at straw polls in the United States, which we see cropping up in the 1800s. So to bolster election coverage, what newspapers would do is they would interview voters as they left polling precincts. At the time, such interviews were called straw polls. And the first recorded one took place in 1824. This became a common practice by local and national newspapers and magazines by the 20th century.
Ken Ward (03:00):
So it wasn’t like this was a social scientific thing from the beginning. It actually just started out as a newspaper practice?
Wendy Melillo (03:06):
It started out really as a newspaper practice evolved more to magazines. And then, you know, it kind of took off by the, you know, start of the 20th century.
Ken Ward (03:21):
So what led to sort of that scientific or social scientific tradition and political polling sort of blossoming in the 20th century?
Wendy Melillo (03:30):
So you start to see at the beginning of the 20th century this concurrent rise in the study of the social sciences in education and government where the public is learning more about things like sociology and statistics. At this time, we also see the beginning of market research firms, which were created to help manufacturers promote and market their goods and services more widely. Right? So at this point, George Gallup enters the picture. He’s one of the first practitioners of scientific polling, and he founded the American Institute of Public Opinion in 1936. We also see a few years later in 1941 the first non-commercial polling agency called the National Opinion Research Center.
And so what happened is Gallup would pull a demographically representative sample to correctly predict, in the case of Franklin Roosevelt’s 1936 reelection win. He would consider polls to be an expression of what the average American thought, not a prediction of election results. And this is the distinction at the beginning because we see that polling is supposed to reflect – the intent of polling was to reflect what the average American thought. And so Gallup calls polling the pulse of democracy in 1940, and he positioned himself as a scientist who publicly proclaimed that he did not vote. He didn’t care about who won elections. He just wanted the polling to be accurate.
Ken Ward (05:09):
Hmm. So did newspaper reporters embrace this new style of polling? Was this something that they were on board with, or did they prefer to stick with their method of straw polling? Was this a popular method of polling that the journalists embraced?
Wendy Melillo (05:24):
Yes. Because you know, they were able to get broader coverage with this. All right. Because, you know, it was labor intensive to do this at every precinct, and polling captured a wider slice of the population. And so you also see at the same time we’re talking around, you know, the 1940s, Elmo Roper is another founding father of political polling. And he agreed with Gallup’s perspective that the purpose was to determine the people’s voice, not sway it. And from the newspaper and magazine perspective, you get more voices into the equation, which was so attractive, you know, to editors at the time. So polls would be the way that the opinions of everyday Americans would be communicated to people in positions of power and the media played a role in that. Right?
Ken Ward (06:19):
Sure. So I see several different, you know, folks involved in this. So if we’re talking about this from this media context, you have the pollsters themselves or the, you know, the people collecting the poll data. But you also have reporters who are using that data and you have politicians who are, you know, their successes and failures are probably represented in this data as well. So you mentioned that Gallup and Roper sort of envisioned polls as just a representation of opinion. That’s how polls are supposed to be used. But in practice I gather it’s a little bit different. So how are polls actually used by politicians and by reporters even in these early years of social scientific polling?
Wendy Melillo (07:00):
Okay. So let’s examine for a moment how the intention to use polling to reflect the opinions of Americans started to shift because that’s essentially then what starts to happen.
Ken Ward (7:16):
Wendy Melillo (7:17):
So we start to see how concerns are raised about how polling results could be used to influence independent voters. And at the time, Gallup dismissed such criticism. He argued that while people think the American people behave like sheep, there wasn’t one bit of scientific evidence to support that conclusion. He argued, right? But this shift we’re talking about from what, you know, Gallup and Roper originally intended, and how then the polling starts to influence American voters. And so scholars start to point out the potential danger that pollsters like Gallup, you know, either ignored or dismissed. So you get to see, you know, scholars coming in saying, well, look public opinion research can be used by politicians to create messages that could be used to target the public. So instead of using public opinion polls to guide policymaking, right, which was, you know, Roper and Gallup’s intent, let’s look at the poll, and here politician, you’re formulating your policy, let the voice of the American people guide you. It ended up shifting. And the polls were then used by politicians to persuade the public to vote in a certain way.
Ken Ward (08:39):
And so here in these early years, was this all theory? You know, that this, these were, you know, a lot of these were concerns of scholars. Was this theory or do we start seeing politicians using the polling data in this way fairly early in this process?
Wendy Melillo (08:53):
We start to see examples of politicians using it this way. So let’s look at an early example is Franklin Delano Roosevelt. He became an early adopter of polling data. So instead of using the public opinion polls to inform his thinking about the creation of policies as Gallup intended, rather he studied the polls to learn how to influence and persuade the public to accept his policies. And there’s a big difference between the two. And you can actually trace, and scholars have done this, there have been studies done where this particular trend can be traced over subsequent presidential administrations. And it, this shift from a desire to understand the public’s policy preferences shifted to politicians, presidents using the polling to promote a certain image and to promote their own personal appeal.
Ken Ward (09:54):
Hmm, interesting. And it’s also interesting to me just as a bit of an aside, that it’s FDR, that administration that seems to capture this, this potential so quickly after this social scientific method of polling comes about. Just because it’s an administration, I don’t know a lot about that administration, but when it comes to other media like radio, right, adopting the fireside chat and finding ways to use emerging media practices so quickly and that administration, it’s just interesting the way that administration tied these things together. So what about reporters then? You know, today, you know, to sort of fast forward to today and then we’ll jump back. You know, the idea of reporters covering the horse race more than the actual substance of an election is something that we all know and understand is kind of for better or worse a part of political coverage. In those early days, is there a lot of horse race coverage or, you know, when does that reporting practice kind of emerge?
Wendy Melillo (10:51):
We start to see — as the shift is happening, we start to see the press then becoming far more interested in using polling to determine who’s going to win an election. And this is where we begin to see the shift reflected in the media coverage because the birth of the horse race idea, you know, is concurrent with this shift. So the press is very interested in knowing who’s going to win an election because that is a reflection of the public. Think about, you know, how a good example is, look at how we watch a sporting event, right? We watch it and we want to know throughout who is ahead, who is winning at each particular moment and time as we watch that sporting event. And then of course we get, you know, the conclusion and we get the winner at the end. And so the public wanted to know because we want to be able to predict the outcome before it happens. And the media echoed this in the reporting and you start to see over time, and this is a criticism that exists till this day where the media is criticized for placing more emphasis on the horse race than on the political candidates policy positions.
Ken Ward (12:19):
And to clarify, is this enabled, is this ability to sort of think that we’re predicting the outcome of the election and the ability to do all of this, does that result from the sampling power that social scientific polling had compared to straw polling? Like, is that where the distinction lies between the old style of journalistic polling and the new mode of social scientific polling?
Wendy Melillo (12:43):
Well, yes, because the sampling takes into consideration a larger swath of voters. So it’s more persuasive, it’s more powerful. And when it’s grounded in social science research and it’s grounded in, you know, scientific analysis, this becomes a very powerful, very powerful persuader because we base this on evidence, right? Evidence is a persuasive tool, but what starts to happen is polling also can reflect the same kinds of problems that crop up when you collect survey data. So for example, when polled, a voter can respond in a way that makes him or herself look good right in the eyes of others instead of act out, accurately reporting who they intend to vote for. And this is how polling then starts to shape voter behavior.
Ken Ward (13:44):
Hmm. So it starts to actually change the way people think and even respond to polling questions?
Wendy Melillo (13:50):
Well, sure. Because there is the whole idea of social desirability. Right? So an example would be if someone asks you, do you intend to vote in the upcoming election?
Ken Ward (14:11):
You have to answer yes. Right.
Wendy Melillo (14:14):
To be socially acceptable. That’s exactly right. Even if you have no intention of voting in the upcoming election.
Ken Ward (14:23):
Interesting. So how did we see some flaws cropping up then in the way polling, you know, the theory of polling operates, but then how it operates in practice? So we see these issues cropping up related to the way journalists use the data, the way politicians kind of manipulate the data, or find ways to use it to their own benefit and the way voters actually sort of adapt to polling practices. So how does this play out? Like, I know just a little bit about this history, and I know that there’s some big failures that happened fairly early on. So what are some of those failures in news reporting that followed from using presidential polls and these ways that these various parties began doing so?
Wendy Melillo (15:06):
Yes. You know, the biggest failure is probably the most famous one that we know of. And we began to see the failures quickly after the Roosevelt years, right? So the most famous example is this Chicago Tribune 1948 headline inaccurately declaring New York Governor Thomas Dewey the winner of the presidential race that year against incumbent Harry Truman. All of the major polling at the time predicted Dewey, so why the polls got the race so wrong is still a matter of debate today. Right? But pollsters like Gallup at the time argued that the polling stopped a few weeks too soon. All right. The thinking at the time was that not much happens in the final few weeks before an election.
Other pollsters thought the problem was really Roosevelt himself because the approach that pollsters took to polling in the 36, the 40 and the 44 elections were all built around Roosevelt. Right? And so people were either for or against him. But in 1948 you had a very different election. You had the frontrunners, Truman and Dewey, but you also had Dixiecrat candidate, Strom Thurmond. You also had the Progressive Party candidate, Henry Wallace. So you end up with a situation where the polls are predicting a Dewey victory of between like 5 and 15 percentage points. And instead Truman wins by about 4.4 percentage points. So what was going on in the final weeks leading up to the election?
Ken Ward (16:56):
Wendy Melillo (16:57):
So you look at it, historically, Democrats were worried about Dewey’s strength in these polls, right, leading up to the election. So the polls have him coming out the winner. This galvanizes the labor vote, right? The labor vote’s energized. Meanwhile, the Republicans felt their candidate, you know, had it in the bag. You know, Dewey had the rig wrapped up. So stay home, play golf, you know, and the fact that they stayed home and played golf actually comes from Elmer Roper’s son, Bud, who later, when he was later interviewed about the race, you know, what he thought happened.
Ken Ward (17:38):
Huh. Wow. Okay. Interesting. So how do these things change as we progress later into the 20th century? Like, do these problems that we see soon after this, this mode of polling rises, do these problems persist throughout the century?
Wendy Melillo (17:50):
Sure. There are consistent examples. So what, you know, let’s just take a more recent one and let’s look at television, right? We’ve talked about newspapers. It’s hard to forget the 2000 presidential election if you were alive at the time. You know, when you think about what happened that night where the TV networks first call the race for Al Gore then for George Bush, then no candidate, and it all comes down to Florida, right? And the polling had Florida for Gore, but the polling was calling the state for Gore before the polls had even closed in the state. All right? Then we went through the whole recount, remember all that dangling chad business?
Ken Ward (18:45):
Wendy Melillo (18:47):
And in the end, the Supreme Court decides the race and, you know, its Bush v. Gore decision.
Wendy Melillo (18:52):
So you can see where you know polling’s not perfect. It, what, you know, what is it really? It captures, it’s a snapshot of a citizen’s point of view at a specific moment in time. But we know time changes, events change, and that voter’s opinion one day something else could happen that affects that voter’s opinion the next day. That’s, you know, that’s what we’re dealing with in terms of the science.
Ken Ward (19:31):
So is there any culpability here on the part of journalists like what, who has the responsibility for these failures when an election poll goes so wrong, especially in, we’re all most interested in these, in these presidential polls. Who, where can we put this responsibility as far as the historical data tells us, where does that responsibility tend to lie?
Wendy Melillo (19:55):
Well, I think you have to look at first the reality of the situation.
Okay. If you’re going to use polling as a predictive device, you are in for a really tricky time. Why? Because we know, just like we know from the sporting event analogy it ain’t over till it’s over, you know, and so the blame should be equally shared. The public wants to know who’s winning. All right. The presidential races are treated like the sporting events, hence the name “the horse race.” But it’s not that simple. The media wants to know who’s winning cause that’s a reflection of what the public wants. And it’s easy for editors to argue that they’re just giving the public what it wants when, in fact, competition among media outlets also plays a role. It’s hard to sit there and say “we’re not going to cover the horse race” when your main competitor has a story on the horse race.
Ken Ward (20:55):
Wendy Melillo (20:56)
You know, I mean, editors don’t want to be in that position. And then when you take into consideration, you know, the flaws in polling, not that it’s polling’s responsibility per se, but people don’t always understand the science behind polling in terms of what exactly it is. That snapshot of a moment in time that I was talking about and put all this together and you end up with the same problems getting repeated over and over again.
Ken Ward (21:24):
Interesting. Well, so that kind of brings me to this question of the present, or the near past. We’ll say it’s difficult to call 2016 history yet. But we’re, you know, with each passing year we get a little bit closer. So there was a study of why polling was so inaccurate in the 2016 election that was done by the American Association for Public Opinion Research. And they effectively found that the polls weren’t really wrong. They claimed that in the 2016 election that the polls fairly accurately predicted the national popular vote in favor of Hillary Clinton, but they found several, like, issues at state level polling and late decision making, which I think is relevant here too, and issues with overrepresenting some populations. So how are we to understand all of this, especially these findings related to 2016 in the context of history, this history of presidential political polling in the US?
Wendy Melillo (22:20):
Okay, well I have to say, first of all, I find it interesting that when pollsters study the failures, the findings often don’t fault the polling.
Ken Ward (22:29):
Ha ha. Of course!
Wendy Melillo (22:31)
So you have to start with national polling can’t, cannot predict the electoral college. All right? And that was the main problem in 2016. It’s expensive to do polling. It really ranges depending on the size and the scope, anywhere from five grand to $250,000 depending on what you’re trying to do. So if we understand this in the context of the history repeating itself, we end up back to where we started our discussion because the disconnect between what impact polling was intended to reflect—the will of the American people—and how polling has come to be used in modern day America—to predict election outcomes—you know, continues. So you have the snapshot, things change, voters can change their mind and the polls may have been accurate at that moment in time, but the conditions are no longer the same.
And this is particularly relevant in the 2016 case because when we talked about some of those other problems with voting that we see, there were Democratic voters in 2016 who just didn’t want to come out to vote because they didn’t want to cast a ballot for Hillary Clinton. And if the polling predicted that Clinton would win, then those voters felt comfortable staying home instead. Oh, my vote’s not going to matter and I won’t hurt anybody by doing this. And the thinking was that by not voting, I’m not going to hurt Clinton because she was, according to the polls, so far ahead of Republican candidate Donald Trump.
Ken Ward (24:22):
Wendy Melillo (24:23):
The point is that polling does affect voter behavior and not always in a positive way. And understanding that influence is probably the most important thing from the history that, you know, we can learn.
Ken Ward (24:41):
Interesting. So we have a, we’re running short on time, but I have one last question for you. And it’s a question that we ask all of our guests. This sort of ties all the episodes together. Why does journalism history or media history matter?
Wendy Melillo (24:55):
Well, I think it’s important to start with a few comments about history in general. Why do we study history? We study it to inform the present and to guide us into the future. Right? And history is now part of the, you know, history is part of the humanities, and we study the arts and sciences because it’s important to do so at a time when these very things have been devalued in our current society. And we face this problem because studying history is considered superfluous and unproductive by some in our society. However, it’s far more critical than that. And the study of journalism history in particular is critical to combating the threats to our democracy. So if you think about journalism history for a moment, don’t we teach our students the watchdog function of journalism? What’s our society going to look like if we do not have a steady source of reliable news?
Do the self-interested politicians dominate with propaganda? You know, how would we understand the complex problems that face us. Right now, with the coronavirus situation, many Americans are looking to the 1918, 1919 influenza pandemic to help them understand the context of what is happening to us right now. And I think we’ve gotten away from teaching journalism history today in the context of the Fourth Estate. I think we get the watchdog function. We do a good, a good job there. But when we think of that term, Fourth Estate, it’s really used to accentuate freedom of the press from government to allow journalists to observe how the government operates, right? And to make sure that politicians and other members of government don’t exploit the democratic process. So I’ll leave you with this final thought. Historically, the press has been instrumental in the birth and growth of democracy. Well, it’s also instrumental in sustaining democracy. And a democracy can’t function without citizens who are informed. And that information that citizens receive must be truthful and factual.
Ken Ward (27:24):
Important ideas for us all to keep in mind, especially in a time like this. Well, thank you very much for being on the show. I really enjoyed our conversation.
Wendy Melillo (27:31):
Thank you so much.
Ken Ward (27:34):
Thanks for tuning in to this episode of the Journalism History podcast. If you like our podcast, please consider leaving us a rating and a review on the podcast app or wherever you listen to us. Until next time, I’m your host, Ken Ward, signing off with the words of Edward R. Murrow: Goodnight, and good luck.