Nonprofit, nonpartisan journalism. Supported by readers.


Bold prediction: almost all House and Senate races will be won by a Republican or a Democrat

REUTERS/Zach Gibson
U.S. Capitol

Some years back, when all the local TV stations were switching from a five-day weather forecast to a 10-day forecast, I asked someone from one of the stations how much confidence he had in the 10th day of the 10-day forecast. Hardly any, he said. So why do you broadcast it, I asked. He said: We all used to do five days, then one stations started doing 10, and we started getting complaints and questions about why we don’t do 10, so we all started doing 10. But don’t put any faith in the 10th day. Even the fifth day often turns out to be wrong.

I often think we spend too much time and effort trying to know what we can’t know, especially about the future and even more especially about political races and how they are going to turn out.

We don’t spend enough time trying to know the knowable. The knowable is mostly made up of things that have already happened, problems and challenges that we actually face, and maybe even the options for dealing with them, individually and collectively.

The unknowable is what will happen in the future. Yes, that’s right. I’m saying that I can’t see the future and I doubt you can either, and that includes the future of all the political races now being contested.

Today, four weeks out from Election Day, we political junkies are all slaves to the freshest batch of poll numbers and, no matter how we try to remind ourselves that that poll numbers are nothing but an out-of-focus snapshot of a not-really-knowable present (considering that the margin of error makes most of the key races technical toss-ups), we still scarf them down as if they were predictions, which they aren’t.

So I laughed out loud yesterday, when I saw headline on a column by Nate Cohn, the New York Times’ political number cruncher that read:

“Battle for the House Has a Wide Range of Possible Outcomes.”

No duh.

I predict that millions of votes will be cast. I predict that almost all of the House and Senate races will be won by either a Republican or a Democrat. All of the self-appointed future-knowers believe that Democrats will make a net gain in the overall House races. None of them know what we all want to know, which is whether the Democratic pickups will be a net gain of 23 or more seats, which would flip control of the House.

Oh, and to be a total jerk about it, might I add that pretty much none of the future-knowers predicted that Donald Trump would beat Hillary Clinton two years ago?

According to the conventional wisdom a week or two ago, the political number-crunchers were issuing a percentage likelihood, and it was way above 50 percent according them, maybe even 80 percent likely, that the Dems would pick up enough seats to take over the U.S. House. And maybe they were right. We’ll never know, for at least two reasons.

Reason one is that the previous wisdom about the future was that Democrats were much more motivated to turn out on Election Day. But now Republican backlash against the Democratic treatment of Brett Kavanaugh has apparently injected enough new wind into Republican sails that those who said it was 70 or 80 percent likely that the House would flip are having a hold-on-a-minute moment. At least that was the reason for the rethink of the past few days.

But the result is that,“Battle for the House Has a Wide Range of Possible Outcomes.”

Poll results have been called “crack cocaine for political junkies.” The point is not that it gets you high. The point is that it gives you the momentary feeling that you can see the future of the political season. Then that feeling blows over and you need another hit. Wash, rinse, repeat.

I guess I’m a political junkie too. I consume great gobs of the latest news (even though I know it will soon be replaced by more news). One of the (many) reasons I like history is that, although it is likely to be subject to constant reinterpretation, at least the facts have already happened.

You can also learn about all our free newsletter options.

Comments (15)

  1. Submitted by jim hughes on 10/09/2018 - 10:25 am.

    Democrats had a lot of momentum a year ago. But they’ve now squandered most of it on an endless MeToo witch hunt, and a too-obvious obsession with check-box candidates. These things really drive right-wing turnout. And what we learned in 2016 – for the umpteenth time – is that polls can’t predict turnout.

    • Submitted by Pat Terry on 10/09/2018 - 10:37 am.

      No, for the umpteenth time, the polls in 2016 were exactly right. The final polls had Clinton 2-3 points ahead of Trump, and that’s how the election went.

    • Submitted by Curtis Senker on 10/09/2018 - 11:59 am.

      Right. And after throwing the kitchen sink in, they lost the battle for the court, then didn’t miss a step before promising impeachment proceedings against Kavanaugh if they win the House. We conservatives could not buy that kind of get out the vote campaign.

      • Submitted by Jim Bernstein on 10/09/2018 - 05:25 pm.

        Nearly all polls show that more people are opposed to Mr. Kavanaugh and the few that don’t show majority opposition show that it is a 50/50 proposition. I don’t think that is much of a hook for conservatives to hang their hat on. Moreover, while Republicans and Democrats are safely in their pro/anti Kavanaugh corners, more independents reject Kavanaugh than support him.

        It also seems that Mr. Kavanaugh’s closet may contain a lot more skeletons than what has come to light. I suspect that Democrats will do a lot more diligence than Republicans ever did to find out just what and how many of those skeletons warrant consideration.

        Conservatives started this Supreme Court war by refusing to even consider Judge Garland who was chosen by President Obama after Justice Scalia died.

  2. Submitted by Pat Berg on 10/09/2018 - 11:24 am.

    So how do we convince people that mid-terms matter? And to just get out and vote, for Pete’s sake!

  3. Submitted by John Webster on 10/09/2018 - 11:29 am.

    Every election I’m amused by all of the hedged, tenuous predictions of the so-called experts. Rarely does anyone flatly predict an outcome for close races; it’s always “this may happen”, “that could be the case”, etc. The latest is that the recent Supreme Court hoopla “could” motivate this side or the other to get out and vote, and it “might” result in one party or the other having a majority in either the House or the Senate. No kidding. Will anyone stick his/her neck out and make a firm prediction?

  4. Submitted by John Evans on 10/09/2018 - 01:23 pm.

    When you’re talking about the future, all you have is evidence from the present and the past. You can only describe future outcomes in probabilistic terms.

    Humans hate trying to think that way; it’s hard, and it’s dramatically unsatisfying. It has to constantly be updated as new data come in. You can never be certain, and our minds crave certainty. Journalists like to talk about certainties. If they don’t have a certainty, then they like to hint at a certainty, and hedge their bets with weasel words.

    But useful forecasting can be done, and is being done pretty well in many places. To use it, you have to train your mind to deal with uncertainty and probability. A responsible journalist would. Almost none do.

    If you’re looking at a weather forecast to determine if it’s going to rain, and the forecast says it will, what does that mean? If you look deeper, the forecasting models actually develop probabilities, not predictions. Weather forecasting models are amazingly sophisticated, the probabilities they produce prove out over time to be accurate. They can’t give you certainties; but they give you the information you need to deal with an uncertain future. In their absence, you would be groping in the dark. If you use them properly, your will save yourself a lot of grief in the future. Your outcomes will be much better.

    The polling, (that is, the data collection,) in 2016 was pretty good. The probabilities that, say Nate Silver’s model produced were, I think, probably accurate. But they were not really predictions.

    Journalists make predictions. In 2016, the journalists were wrong. Journalists love to blame pollsters and forecasters for their own irresponsible reporting. And that’s the most irresponsible and juvenile reaction possible.

    Eric, I’m disappointed in you.

  5. Submitted by Paul Brandon on 10/09/2018 - 07:14 pm.

    That’s better than Roth waxing Jennings, as Groucho would say.

  6. Submitted by Paul Brandon on 10/09/2018 - 07:20 pm.

    For any mathematicians in the audience, there’s a field of statistics called Bayesian Analysis (derived from Bayes Theorem formulated in 1763), which takes into account previous knowledge in making predictions, rather than assuming that everything is 50/50.
    What this means in regard to political predictions is that depending on our assumptions, we may predict different outcomes for the same events. Since these assumptions are not verified facts, different analysts will make different predictions for the same event. That doesn’t mean that some of them are smarter (more competent) than the others, just that in hindsight their assumptions were more accurate.

    • Submitted by Pat Berg on 10/10/2018 - 07:48 am.

      Is there any straightforward way to tell whether or not any given analyst is using Bayesian Analysis (vs. the alternative(s))?

  7. Submitted by Kent Fralish on 10/10/2018 - 11:44 am.

    I believe it goes, Lies, Damn Lies and Polls….

    • Submitted by Paul Brandon on 10/10/2018 - 02:43 pm.

      Sam Clemens’ phrase was ‘lies, damn lies and statistics’.
      Polls are a method of prediction based on conditions at the time of the poll.
      Circumstances may change even in the days before an election (take Comey’s statements, please).
      Since the standard error of measurement of most polls is in the 4% range, they are necessarily of limited value in a close race. Predicting that the odds of someone winning are 4 to 1 is not a guarantee; unlikely events do happen, which is why they are likelihoods, not certainties.

  8. Submitted by Paul Brandon on 10/10/2018 - 12:36 pm.

    Some analysts such as 538 publish some info on their methods — I’m not sure that Nate Cohn has.
    Most people regard their algorithms as trade secrets and don’t publish the details, but I would assume that at least some Bayesian principles are included. That would be standard practice these days.

  9. Submitted by Paul Brandon on 10/10/2018 - 12:37 pm.

    That should have been Nate Silver.

Leave a Reply