• mozz@mbin.grits.dev
    link
    fedilink
    arrow-up
    29
    arrow-down
    2
    ·
    5 months ago

    The polls are absolute shit. You can draw a little bit of a conclusion that if Biden went up by 2 points, that probably means that wherever the real answer is, it might not be crazy to think it went up by somewhere from 1 to 3 points because of something that happened. That part may be worth being slightly happy about. But whether that real answer is +2 like they say, or -10, or +20, modern polls actually can’t tell you, and all these people that are telling you they can, are lying. In my opinion.

    Source: I looked through the methodology they actually use for this polling and found it to be dogshit, and then looked up a few polls for recent elections and found that the poll differed from the actual result of the election by an average of 16 percentage points.

    • dragontamer@lemmy.world
      link
      fedilink
      English
      arrow-up
      19
      arrow-down
      1
      ·
      5 months ago

      Polls are dogshit.

      But poll movements are worth tracking as long as the dogshit quality remains consistent.

      • KevonLooney@lemm.ee
        link
        fedilink
        arrow-up
        8
        arrow-down
        1
        ·
        5 months ago

        True but what the poster said is important: polls do not work well in close races because they don’t sample correctly. The average poll respondent is older, whiter, and more conservative than the average voter. So you end up with a skewed sample. Plus you sample so few of the other voters that you can’t make a guess as to what they think.

        No younger people answer the phone for pollsters, so this is hard to fix. They are probably going to have to start paying people to answer surveys. I know I won’t answer them for 15 or 20 minutes for free.

        • dragontamer@lemmy.world
          link
          fedilink
          English
          arrow-up
          2
          ·
          edit-2
          5 months ago

          True but what the poster said is important: polls do not work well in close races because they don’t sample correctly.

          That’s not my point. My point is that the poll movements are almost always correct, because pollsters are at least consistent with how biased they are / errors in sampling.

          Fox News had Trump leading Biden a few weeks ago. Today, Biden is ahead of Trump. We don’t know where the “truth” lies, but we can 100% conclude that the typical American has lost a bit of favor on Trump in the past few weeks.

        • spongebue@lemmy.world
          link
          fedilink
          arrow-up
          2
          arrow-down
          1
          ·
          5 months ago

          The average poll respondent is older, whiter, and more conservative than the average voter. So you end up with a skewed sample.

          Any reputable polling group will adjust for that. Granted, fewer and fewer people are answering their phones and taking these polls, but basic demographics are a well-known and pretty easy to adjust for thing. Most polls take a lot of that information for that reason

          • JimmyMcGill@lemmy.world
            link
            fedilink
            arrow-up
            1
            ·
            5 months ago

            Yes

            People here and on Reddit really think that polls are made with just a few calls and then some average/extrapolation and that’s it.

            Meanwhile it’s an entire field with a lot of complex math and people with more knowledge about it than everyone in this comment section combined.

            And then the classic “polls are shit, they always get it wrong”. By definition polls are correct because they just represent an objective data set. Then they translate it into a phrase that we humans can (somewhat) understand but people then take it wrongly.

            They read “Poll X says candidate Y will win” when instead they should read “According to the data obtained for Poll X, candidate Y has a Z% chance of winning with a confidence level of W%”. And that isn’t wrong unless someone wants to find the mistake in the math.