Political polls are addictive — and don’t tell the future

According to political (and sports) statistics maven Nate Silver, as of four minutes ago, Hillary Clinton was 79.6 percent likely to win the presidential election, compared to Donald Trump, who was 20.4 percent likely to win. The full workup is here.

I’m not an expert on who is an expert on this, but I do believe Silver has established himself as one of the tops.

I assume this percentage breakdown makes some of you feel better about things, others not so much. For me, it makes me want to warn against thinking that anyone — even Mr. Silver — can see the future in matters like this, nor does he claim to. I’m sure he can justify his numbers, but if you actually listened to his justification, you would conclude that he isn’t making a “prediction,” he’s making an estimate of likelihood, based on things he thinks he knows, and leaving room (but, in this case, not a whole lot of room) for the possibility that things he doesn’t know will happen and will change the outcome, or not even really “change” the outcome, since it never happened, but bring about a different outcome from the one he thought most likely back in June.

I look at poll numbers quite often, much more often than is healthy. I can’t help it. They are something like drugs to an addict. I try not to write about them too much, not only because I know they don’t really tell the future and can only estimate the present within a margin for error so large that in most cases they don’t even really claim to measure who is ahead.

This is more than just statistical pettifoggery. History — even the relatively short history of polling — is rife with presidential candidates who were behind by more than Trump is behind in current polling and nonetheless won. But also, the time you (and I) spend obsessing on the latest poll numbers is time we could be spending with our friends and families, helping our communities, working for remuneration or even deepening our understanding of how our dear ol’ country and our world got to the place that it is in right now and how it might get to a better place. I highly recommend any of those alternatives.

And now, if you can spare me for a minute, I’ll just take a peek at the just-updated, Real Clear Politics log of the latest polls. OMG. The top one, by Rasmussen, has Trump ahead by 4 points. The next one, taken in the same country at pretty much the same time, by Public Policy Polling, has Clinton up by 4.

You can also learn about all our free newsletter options.

Comments (17)

  1. Submitted by Paul Brandon on 06/30/2016 - 07:50 pm.

    31 to 4

    Looking at the first page of the Real Clear Politics log (going back a week or so) Clinton was ahead in 31 polls; Trump in 4. This answers my question about which polls are outliers.

  2. Submitted by Hiram Foster on 07/01/2016 - 07:31 am.

    Margin of error

    “I try not to write about them too much, not only because I know they don’t really tell the future and can only estimate the present within a margin for error so large that in most cases they don’t even really claim to measure who is ahead.”

    Something to understand about “margin of error” is that it has literally nothing to do with the margin any given poll might be in error. Margin of error in polling theory has to do with the variation in polling results, not their accuracy. It’s sort of like shooting an arrow. You might not ever get close to hitting the target, but if you hit the broad side of a barn with regularity, the margin of error is just fine.

  3. Submitted by Neal Rovick on 07/01/2016 - 08:10 am.

    Look at the Brexit polls.


    What seemed certain 4 months before the vote was found different at the vote.

    It’s like looking at the weather report 7 days out.

    Calm assumption of rationality can be your enemy.

    • Submitted by Theo Kozel on 07/01/2016 - 10:29 am.

      Already understood

      Yes, Nate Silver and the entire gist of this article quite clearly and explicitly address this point. There is a lengthy discussion on the 538 website about what this ‘80% probability Clinton wins’ really means. Silver and company have repeatedly and consistently written qualifiers about what probabilistic modelling can and cannot do, and the possibility of a Trump win is very much covered by that.

      Misunderstanding the meaning of statistical modelling can be your enemy.

      • Submitted by Neal Rovick on 07/01/2016 - 01:32 pm.

        I understand polling caveats, but the absurdity of a prediction four months out by anyone other than Hari Seldon is enormous. And given that the probability of winning in the end can vary by each news cycle and revelation, this type of poll is especially dubious. The “if the election were today, who would you vote for” poll has more validity than this type of meta-poll crap. In the end, what testable proof of the method is there ? Either Clinton is elected or not ? Was she 60% close to being elected ? Would Trump have beaten the 30% odds ? What do those statements actually mean ?

        Let’s say Trump actually did go out and shoot someone in the street….


        Let’s say that a photo turns up with Clinton running toward the benghazi compound waving the black flag of ISIS…

        It’s a perpetual job for Mr. Silver.

  4. Submitted by Neal Rovick on 07/01/2016 - 08:42 am.

    Of course a red-hot VP choice like Christie or Gingrich will swing the polls…

  5. Submitted by Hiram Foster on 07/01/2016 - 09:55 am.

    The Brits are notoriously difficult to poll. I think the problem is that they just don’t know how polls are supposed to work. Instead of responding to them naively, I think they try to manipulate them.

  6. Submitted by chuck holtman on 07/01/2016 - 10:57 am.

    It’s the neighborhood crank again

    When was the last time we as a society, and our elected leaders, had a half-cogent discussion on an important matter of policy, and the media supported the discussion by providing useful information? I can’t remember.

    In the interests of profit and power (as always), over time self-government has been degraded to a spectator sport. Cheer for your side, hurl insults at the other side, and when the game is done, move immediately to the next. The horse race is 98%, what the candidates actually do once in office is 2%, and those with the power and the media shares like it that way just fine.

    Polls are just one more product marketed to a pacified citizenry and express the degradation of civic engagement.

    • Submitted by RB Holbrook on 07/01/2016 - 03:50 pm.

      As One Crank to Another . . .

      I agree with what you’re saying. However, if the media is driven by profits (ratings), the horse race coverage they are putting out in some way reflects what the public wants to see. It’s a vicious circle. Lofty protestations to the contrary, most people prefer watching the “exciting” coverage of where did a fight break out to any real substantive discussion. The media obliges–after all, they’re not in this for their health.

      It is a point of pride for many that they “don’t care about politics.” They vote, but it’s based on who seems like the nicer guy or whose rallies are more like WWF than C-Span. Those masses are the ones who are voting with their TV remotes, and dictating the coverage the rest of us see.

      • Submitted by chuck holtman on 07/03/2016 - 08:56 am.

        Yes, the blame falls on the populace.

        If you don’t care to exert the effort to be informed and thoughtful, to take your civic role seriously for its impact on others and on the future, you can live your life but you shouldn’t vote and you shouldn’t voice opinions as if they deserve some consideration.

        But the blame also falls on the establishment media. A pimp isn’t morally absolved just because men want prostitutes. Each time I walk past a media screen and see yet a more egregious advance in distraction and pandering, I wonder how the folks involved, from the corporate execs to the talking heads, can look at themselves in the mirror.

        I guess “moral scold” is more accurate than “crank.”

        • Submitted by RB Holbrook on 07/05/2016 - 10:35 am.

          Plenty of Blame to Go Around

          “But the blame also falls on the establishment media. A pimp isn’t morally absolved just because men want prostitutes.” No, but the pimp would go away if no one were willing to pay for his services. It’s capitalism writ large.

          I’m old enough to remember when the media gave us news we were supposed to know. The big story of the day might be about a coup in South America, or a sharp downturn on Wall Street. Boring, but essential, and that was a viable business model when there were three TV networks and one or two newspapers to inform us. Now, with a 24-hour news cycle and no end of news outlets, the competition is cutthroat. Viewers who are bored with a story about how the weather will affect corn and soybean crops can turn to another station and watch a “news” report on where to buy the best hot dog buns in the state. News outlets will deny with one breath that they are “ratings driven,” but then shrug their collective shoulders when asked about the fluff they purvey and say “it’s what the people want.”

          American society has increasingly decided that nothing is worth doing unless it can be monetized. What would happen if a TV network or station (still where most people get their news) started devoting all of its programming to substance? How long would it stay in business?

          • Submitted by Jim Million on 07/07/2016 - 02:05 pm.

            Not Long, for Sure

            Broadcast TV cannot pose “$64,000” questions when the average viewer’s decision point is somewhere around $29.95.

  7. Submitted by Hiram Foster on 07/01/2016 - 11:29 am.

    Election coverage

    Elections don’t take place in campaign, they take place among voters. I think something to try is for the media to stop covering campaigns. Just use pool coverage and Cspan to cover speeches and events. The reporters should then be reassigned to talk to voters and to actually cover stories.

    My modest proposal is that reporters and journalists should not be allowed inside the District of Columbia except when Congress is in session. When the congress types go home, reporters should follow them and cover what they do in their constituencies.

    • Submitted by Jim Million on 07/07/2016 - 02:13 pm.

      Too Sensible an Idea

      The average reporter could then no longer easily get the “news” from other reporters.

      As for poll reporting: Ask any “talking head” to define statistical validity as opposed to reliability.

      An average of Polls is just some kind of sausage, isn’t it?

  8. Submitted by Paul Brandon on 07/02/2016 - 07:56 pm.

    In fact

    They don’t even tell the past
    (think about it).

  9. Submitted by Hiram Foster on 07/05/2016 - 09:18 am.

    The future

    Polls assume that you will do what every investment prospectus you ever read will warn you against in the stern terms, that past history isn’t a predictor of future performance. Pollsters really need to assume that it is but paradigm busting candidates like Donald Trump pretty much thwart that. We simply don’t know what kind of voter pool a Trump candidacy will attract, and therefore our ability to predict what voters will do. In addition, Trump being Trump, we don’t know how people questioned about him will react. Trump is, for example, widely perceived as a racist. One thing this means is that many people who will vote for Trump will be reluctant to associate themselves with him publicly. We see this most visibly in the lengthening list of Republicans who are finding reasons not to attend their party’s convention. While lawn signs don’t play much of a role in presidential campaigns in Minnesota, I don’t expect to see many lawn signs supporting Trump for the simple reason that people don’t want to be publicly associated with him. But that doesn’t at all mean that they won’t be voting for him.

  10. Submitted by Tom Regnier on 07/22/2016 - 09:42 am.

    Polls are not predictors of the future. Polls are a snapshot in time of the answers provided by a specific sample of individuals. Using proper statistical polling methods, the results can often be projected to represent a larger universe.

Leave a Reply