9 votes

Republicans win two U.S. House special elections

9 comments

  1. envy
    Link
    When Trump accuses the Democrats of something they never did, it's usually him projecting. When Trump accuses the Democrats of something that no one has done (yet), it always alarms me. I fear the...

    Despite focusing on the supposed unreliability of mail ballots, Trump expressed additional outrage when the state of California agreed to add a polling station in the city of Lancaster, bringing to 13 the number of places where voters could cast ballots in the congressional district (as opposed to 1,000 in a normal non-coronavirus election). Despite Trump’s tweet that “They are trying to steal another election. It’s all rigged out there. These votes must not count. SCAM!,” the additional polling place was actually requested by the Republican mayor of Lancaster.

    When Trump accuses the Democrats of something they never did, it's usually him projecting.

    When Trump accuses the Democrats of something that no one has done (yet), it always alarms me.

    I fear the Republican party are planning to severely limit polling places in Democratic strongholds, while blocking vote by mail, and will blame Covid19.

    9 votes
  2. [7]
    gpl
    Link
    I wouldn't read too much into these results as indicators of enthusiasm come November; nonetheless, this isn't a great outcome for Democrats. I don't think this is an indication of a #redwave or...

    I wouldn't read too much into these results as indicators of enthusiasm come November; nonetheless, this isn't a great outcome for Democrats. I don't think this is an indication of a #redwave or anything remotely close, given that they are special elections during an epidemic whose seats will be up for grabs in a few months. Completely anecdotally though, it does feel like there is less enthusiasm for the November elections than at a similar time before the 2018 midterms.

    5 votes
    1. [6]
      goodbetterbestbested
      Link Parent
      Here's a project no one is doing: I haven't seen any "election prediction" sites taking the polls from 2016, comparing the final results to the final polls, and then applying the difference...

      Here's a project no one is doing: I haven't seen any "election prediction" sites taking the polls from 2016, comparing the final results to the final polls, and then applying the difference between those two things to current polls in order to predict what the final results might look like state-by-state. That is: using the "error" in the polls from 2016 to "correct" the polls for 2020. It wouldn't be a crystal ball or anything but it would give us a baseline to temper our expectations.

      2 votes
      1. [5]
        gpl
        Link Parent
        That's because the polls were about as good as average in 2016. Trump beat his polling predictions by a couple of points in the states that mattered, but the end result was by and large still...

        That's because the polls were about as good as average in 2016. Trump beat his polling predictions by a couple of points in the states that mattered, but the end result was by and large still within the error bars given by those polls. This wasn't really reflected in the national polls because, while polls underestimated Trump support from white, non-college educated working class voters, they also underestimated Clinton's support from non-white, working class voters. On a national scale these canceled out so the national polling was pretty good, whereas on a state-by-state level the error bars were bigger. Another issue leading to the perception the polls failed or were systematically off in 2016 is the fact that polls since 2000 have actually been a bit better than average if you count polls going back to the 70s. So we were primed to think polls were actually better than they have been, on average.

        All of this means the errors in 2016 fell within the quoted error bars in almost all cases, so I don't think it would be statistically sound to 'correct' current polls by shifting those error bars around.

        10 votes
        1. [4]
          goodbetterbestbested
          Link Parent
          Sure, I'd get an F on this paper in stats. I'm also not saying that 2016 was a uniquely bad year for polling, I'm not one of those. But I think the question of "What if the polls are off by the...

          Sure, I'd get an F on this paper in stats. I'm also not saying that 2016 was a uniquely bad year for polling, I'm not one of those.

          But I think the question of "What if the polls are off by the same amount as in 2016? What is Biden's current position in that case?" is not a worthless one. It might even be just as worthy as the question, "What does Nate Silver's extremely complex model tell us the chances are?"

          2 votes
          1. [3]
            skybrian
            Link Parent
            The bottom line is that predicting the future is hard and we need to learn to live with uncertainty. We won't know anything for certain until after the election. There's no important reason for us...

            The bottom line is that predicting the future is hard and we need to learn to live with uncertainty. We won't know anything for certain until after the election. There's no important reason for us to pay attention this early; it's just for entertainment.

            But given that, I'll still pay more attention to election models like 538 than to other guesses, because I trust them to formally keep track of how well they're doing and make adjustments.

            5 votes
            1. [2]
              goodbetterbestbested
              Link Parent
              Sure, but don't you think assuming that the polls will be off in roughly the same way as the most recent presidential election is a good pessimistic starting point? Maybe pessimistic isn't the...

              Sure, but don't you think assuming that the polls will be off in roughly the same way as the most recent presidential election is a good pessimistic starting point? Maybe pessimistic isn't the right word, I'd say the error of the last election's polls just seems like a pretty important thing to consider in a prediction.

              2 votes
              1. skybrian
                Link Parent
                Yes and no. Yes, when there is systematic error like this then it's easy enough to correct for it. For example, 538 keeps track of bias for each polling company and they will correct for that in...

                Yes and no. Yes, when there is systematic error like this then it's easy enough to correct for it. For example, 538 keeps track of bias for each polling company and they will correct for that in their model.

                But I'm not sure that adding an additional bias on top of what they do already would add to their accuracy; it might just be double-counting? Someone would have to do the math to figure out if there is any gain to be had, but I doubt it's that simple to improve on 538's models or they would fix it themselves.

                Since I haven't done the math and have no special knowledge about elections, I will probably just take their predictions as given, once they start making them. (It's still very early.) I might look at other election forecasters, but I don't have any good criteria for choosing between them.

                That's still paying more attention than is really necessary. I'm a news junkie so I do it for fun, but really, having an early guess about November election results doesn't matter to any decision I'm likely to make.

                5 votes