miguel balderrama's home timeline

Notices

  1. Matt Matt

    That probably doesn't happen today. Elections today are very partisan and correlated and boring and if a Presidential candidate wins in a given state then other people from his party on the same ballot also win statewide - that hasn't quite begun to happen with individual congressional districts though, Clinton won a great deal of congressional districts that ultimately voted for Republican congressmen (though most of those people lost their re-election bids in 2018).

    Monday, 23-Sep-19 03:57:08 UTC from web
    • Matt Matt

      In the case of 2008, Obama actually underperformed the incredibly lofty 9 point advantage the generic ballot gave the party, but the party as a whole did incredibly well and won districts and states that Obama himself could not win due to ballot splitting.

      Monday, 23-Sep-19 03:54:42 UTC from web
      • Matt Matt

        Pretty much no causation and only correlation, but errors in the generic ballot tend to result in higher swings towards Democrats, rather than the opposite. That's just a trend I've noticed that doesn't actually point to anything greater, except maybe that Democrats respond to congressional polls less often than Republicans do since they usually only care about the Presidential Elections.

        Monday, 23-Sep-19 03:51:26 UTC from web
        • Matt Matt

          ...you can, pretty reliably, predict which states would go what direction supposing the election happened today. http://rainbowdash.net/url/875208

          Monday, 23-Sep-19 03:48:39 UTC from web
          • Matt Matt

            If we wanna talk poll accuracy, I find that FiveThirtyEight's generic congressional ballot works pretty damn well on a national scale. 4 of the last 5 Presidential elections had their national popular votes fall within two points of the final generic ballot margin - including 2016. Now, if we apply that to state calculated partisanship, using either the Cook PVI (more friendly to Democrats, as they only take the 2012 and 2016 Presidential elections - two very neutral but still otherwise slightly Democratic leaning years - into account), or the 538 PVI (more friendly to Republicans, as it also accounts for more local elections between 2012 and 2016, including the Republican wave year of 2014) as our baseline... http://rainbowdash.net/url/875207

            Monday, 23-Sep-19 03:45:30 UTC from web
            • Matt Matt

              But that's not even the point I was making. As a more conservative Democrat, Biden is objectively more likely to win over Republican-leaning voters than somebody like Elizabeth Warren.

              Monday, 23-Sep-19 03:38:54 UTC from web
              • Matt Matt

                What people don't get about polls, and what they especially seemed to not get after Trump was elected, is that capturing 51% in a poll is just that - a 51% chance that you win. Clinton's lead in key states was always narrow and the fact that she underperformed some of them (and the fact that she overperformed a lot of others, like in Texas, Georgia, and Arizona, which were a lot closer than most pollsters predicted they would be) doesn't make the poll "wrong". It just means the less-likely thing happened, and as it turns out, unlikely things happen pretty frequently. Statistics!

                That's why you don't want to have a two-point lead in a statewide poll, as Clinton did in several, and as Trump currently has in some SUPER red states. You want to have a ten-point lead, as Biden does in several - and even that doesn't create a guaranteed win, it just creates a more likely win.

                Monday, 23-Sep-19 03:38:24 UTC from web