Nonprofit, nonpartisan journalism. Supported by readers.


Why we shouldn’t care so much about polls — even if we can’t stop looking at them

REUTERS/Brian Snyder
Gov. Scott Walker and Donald Trump chatting during a commercial break at the Aug. 6 Republican presidential candidates debate.

First a word about why we should pay less attention to polls and then some fresh poll numbers and finally a little historical anecdote.

Political junkies like me can’t help looking at poll stories. We should spend less time on them and we should definitely learn to treat them as something less than actual information worth knowing, kind of like the 10th day in the 10-day weather forecast, which have about as much chance of being right as a “fact” told by Michele Bachmann.

We were told, based on polls, that Scott Walker was leading the huge Republican field in Iowa and therefore had to be comsidered as among the serious candidates for the nomation. He soon became the first candidate to drop out of the race. Was that because the polls were “wrong.” Not really. It was evanescent. It was based on asking people who had not really formed a firm intention to even attend a caucus let alone whom they would support if they did what they would do six months later if they happened to go and hadn’t changed their minds about whom to support.

When Donald Trump first emerged as the poll leader in the Republican field, Chris Cillizza of the Washington Post wrote a piece headlined “Why no one should take Donald Trump seriously in one simple chart.” The simple chart showed that, although Trump was then on top of the field when Republicans were asked who was their first choice for the nomination, only 23 percent of Republicans viewed him “favorably” compared to  65 percent of Republicans had an unfavorable view of him. As the field narrows, Cillizza reasonably argued, Republicans will be redistributed among the remaining candidates and they will choose from among those whom they view favorably. Trump had no upside because almost everyone who didn’t already support him viewed him unfavorably. Reasonable, right?

Sort of, but not really, as Cillizza honorably conceded in a follow-up post five weeks later headlined “Boy was I wrong about Donald Trump. Here’s why.” The short version of “here’s why” is that Trump’s negative number among Republicans had dropped from 65 percent to 40 percent and the portion that viewed him favorably had shot up from 23 to 57. If your “ceiling” is 57 percent of Republicans, you obviously might end up as the nominee. The first poll asked people who hadn’t really formed an impression of Trump. I wouldn’t want to oversell the bankability of the second poll either. But it was enough to induce Cillizza to bravely admit error. (Personally, I’ve taken a vow to stop speculating on how the Trump story will turn out.)

Pardon the sports analogy but: The score at the end of the second inning of a baseball game doesn’t “predict” the outcome of the game. But at least it’s a real score. The runs that have been scored to that point will stay scored. The runs can’t change their mind and be transferred to the other team. You can’t say that about poll respondents’ current feelings about a candidate in a far-off contest.

But we — at least us political junkies — can’t stop looking. The guys in charge of putting out fresh political “news” at NBC, which polls in partnership with the Wall Street Journal, led today’s edition of “First Read” with “Bad News for the GOP and for Hillary Clinton.” It begins:

“With just over 13 months until Election Day 2016, the Republican Party should be sitting pretty. The Democratic frontrunner, Hillary Clinton, has seen her popularity drop considerably in the new NBC/WSJ poll. What’s more, 62% believe the country is headed in the wrong direction, which isn’t good news for the party holding the White House. And President Obama’s approval rating — at 47 percent — isn’t bad, but it’s not great, either. But here is the challenge for the GOP: Almost every other indicator in the poll is a real problem for the party. For starters, the Republican Party (29%-45% fav/unfav rating) is significantly more unpopular than the Democratic Party (41%-35%). Its current presidential frontrunner, Donald Trump, is the most unpopular figure in the entire poll. In addition, GOP voters are in revolt against their own leaders: 72% are dissatisfied with House Speaker John Boehner and Senate Majority Leader Mitch McConnell (which helps explain why Boehner is stepping down from his job). And Republicans are in a MUCH different place than Democrats and independents on issues like abortion, gay rights, and race.”

Lastly, when I was working up a recent piece about the history of presidential elections, I meant to include a section on polling (but it got too long, so I dropped it). But I was curious about the most famous “polling disaster of all time,” the polling on the 1948 Harry Truman-Thomas Dewey race that lead to the famous picture of victorious President Harry Truman holding up the Chicago Tribune with a banner headline “Dewey Defeats Truman.”

There were several reasons for the blunder, but one of the big ones was that the professional pollsters generally believed that few voters changed their minds in the last two weeks — about whom they supported or about whether to vote. As a result, Gallup, for one (but I gather this was true of other polling operations), stopped sampling 10 days before the election.

Now (as one who has sat in a newsroom on many Election Days awaiting results would know) the polling never stops and voters leaving the polls are asked for whom they voted to see how it matches with the latest round of phone or internet polling.

You can also learn about all our free newsletter options.

Comments (9)

  1. Submitted by Hiram Foster on 09/29/2015 - 02:13 pm.


    There are many more recent failures of polling than the Truman Dewey race. There are a number of problems with polling. To begin with, only about 20% of those polled are actually responsive. The result is that in any given poll, we don’t know how most people feel. Pollsters routinely give margins of errors for their polls. Since they don’t have definitive numbers to compare to, how could they possibly know what the error of their polls could be?

    In ballpark terms, I think most polls you see in the media are not terribly inaccurate. But what also seems to be the case is that polls will fail, when they are most critical. That happens when voters decide late, or actual turnout differs from polling models.

    In ethical terms, what bothers me is that polls are not found news, they are paid for news, a form of checkbook journalism. News organizations buy them, and then expect their reporters to cover them on the assumption that they are accurate. Business considerations are determining news decisions, and I think that’s wrong.

  2. Submitted by Neal Rovick on 09/29/2015 - 05:18 pm.

    Polls create a whole new layer of media stories. The numbers are stories, the changes in the number are another set of stories, and the reactions to the numbers are stories in themselves. And the subdivision by nation, region, state, city, race class, etc. It’s endless….and you can rinse and repeat every week.

    This is especially helpful in the periods when the candidates are essentially doing and saying the same thing over and over at countless stops. It provides a steady check for a lot of people.

  3. Submitted by Hiram Foster on 09/30/2015 - 06:51 am.

    Media stories

    “The numbers are stories, the changes in the number are another set of stories, and the reactions to the numbers are stories in themselves.

    The problem why have is that the news media should be covering, not creating news. That’s particularly true when, as with polls, the media can provide us with no convincing evidence that the news they pay to create, has any sort of objective reality. Sure there is a lot of down time in campaign coverage. But it would be a lot more valuable to the voter if the news media spent it’s energy and money on actually covering what candidates say, instead of the fabrication of polls whose meaning is unclear, if they have any meaning at all.

  4. Submitted by Hiram Foster on 09/30/2015 - 08:01 am.

    A modest proposal

    One of the conventions of political reporting is the boys and the girls on the bus thing, the fact that political candidates are trailed by masses of reporters who are largely given nothing to report. For me, the most amazing scoop from the 2012 campaign came from the guy who secretly recorded Romney “47 percent” speech before a group of donors. Isn’t it incredible that with all the time, effort and expense media organizations spend on following candidates, the only memorable story of the presidential campaign came from a bartender with an iphone set to record?

    Here is my proposal. Instead of each news organization following these guys around, they should reduce the coverage to one pool reporter whose job it is mainly to report if the candidate says anything foolish. The rest of the reporters should report on the reals campaign. For example, there is the political money beat. We are told that campaigns raise and spend huge amounts of cash, because it says so in various filings. But who contributes the cash? What is their agenda? What do they expect to get in return for their money? Who was on the guest list at the Romney function and what did they want? How is the money spent? These are critical questions for anyone who wants to know what goes on in a political campaign and they go almost entirely unanswered.

  5. Submitted by Neal Rovick on 09/30/2015 - 08:04 am.

    Polls are the easiest stories for journalists to write-the who what where when are all in the numbers, the why and how are obtained by a quick call or two to the usual experts. And there is no way that the journalist has any risk exposure in the story–it’s all opinion of others.

  6. Submitted by Jon Kingstad on 09/30/2015 - 08:37 am.

    Polls are one way

    for people to know what other people might be thinking and what they themselves should be thinking. It says something about public opinion and which way it is blowing. It’s also a way for the media to let people know that there is a race on and something “important” is going on.

    • Submitted by Paul Brandon on 09/30/2015 - 09:38 am.

      Thought vs Action

      The problem is that polls record what people are willing to say to someone else at a given point in time.
      Their correlation with what people will actually DO at some later time is weak.
      At best they are a snapshot when we need a real time movie.

  7. Submitted by Hiram Foster on 09/30/2015 - 08:53 am.


    I have no problem with easy to report stories. What bothers me is that so often poll stories are written and published, not because of the independent judgment of the news reporters and editors, but because corporate pays for them and expects to see them covered in the newspaper. And in my view, the stories are often miswritten. A typical news story covering polls says “50 percent of Minnesotans believe such and such”. Well the reporter knows no such thing, at least not from a poll. What the reporter knows at best is that in a sampling of a few hundred individuals the pollster claims is representative of some group of Minnesotans the sample claims to represent shows whatever it shows. Often polls tell us really weird things, like some huge percentage of Republicans believe that the president was born in Kenya or the moon maybe. Nobody ever seems to consider that the respondents are pulling the pollsters legs.

    I have talked about the problem of margin of error. No one can ever tell how much something is in error unless we have correct results with which to compare to. Pollsters are hardly ever in a position to compare to correct results, so despite their claims they are hardly ever in a position to tell us how much their polls might be in error. But in one small area, we do get a clearer sense of what’s going on. Once in a great while, we see after election polls when voters are asked who they voted for. What these polls find is that the polling results often differ significantly from the actual election results. These discrepancies are dismissed as a sort of front running by voters, who associate themselves with winning candidates whether they voted for them or not. But how do we know that? Remember people who respond to polls are always a self selecting group, and that the vast majority of people who are polled refuse to answer the pollster’s questions. We don’t know what those unresponsive individuals think because they don’t respond. But what if it’s the case that loser voters are more unresponsive to after election polls than winner voters, accounting for the discrepancy? And if this unbalanced responsiveness affects after election results, why wouldn’t we think they affect all polls all the time?

  8. Submitted by Roy Sorenson on 09/30/2015 - 10:14 am.

    Walker first?

    I thought Rick Perry was the first Republican candidate to drop out on Sept. 11, Scott Walker on Sept. 21.

Leave a Reply