If the polls are 2018 levels of accurate (within half a point) and current trends roughly hold this is more or less what the election is going to come down to. (And Nevada is completely worthless).
If the polls are 2018 levels of accurate (within half a point) and current trends roughly hold this is more or less what the election is going to come down to. (And Nevada is completely worthless).
Across both 538, RCP, and a few other reliable polling sites as of late the general overall trend is- North Carolina, Georgia, and Arizona all go red, with the former weakening the most as it got the least investment. (his checks out as the third party balance nationally has shifted to be less hostile to Republicans. Get rid of third parties completely in 2020 for both sides partitioning the voters and Trump wins Georgia and Arizona then too)
Michigan has been very strongly blue, strongest in 2020 and second strongest in 2016(Nevada is slowly trending Red so ignore that). Wisconsin was super swingy the last two elections, having extremely bad polling and being the reddest of the rust belt both times. However, Tim Walz strengthens this state more than any other while losing Biden and not picking Shapiro weakens Pennsylvania more than most, so barring another massive upset it's going to be bluer than PA, solidly blue in most polls.
Nevada and Pennsylvania are the swingy states. Nevada has a slow weak red trend, Pennsylvania has had a ton of investment and stung from the Biden dropout. Nevada might have mattered in the Nebraska Law Change scenario, but without that it's worthless. Both have had tight polling for a while, albeit Nevada has more consistently leaned blue while Pennsylvania leaned red for a bit pre-debate.
Of course the polls could be wrong again. A 2022 style error and Democrats sweep the swing states and maybe pickup a pink state. A 2020 style error and everything not Michigan falls Red. 2016 level error means Michigan and Virginia too. But I don't see it happening. They've had two national elections to correct for Trump. They've had one big election post-Dobbs and several smaller ones to correct for that error(which was smaller than the Trump errors and made in the shadow of Post-2020 poll corrections). This is the first time both those factors are going head to head nationally and the pollsters have had a chance to weigh both of them. I don't expect badly wrong polls. But just a half a point off determines the election. Being dead on correct right now favors the democrats, but it didn't the day before the debate. It could go either way.
You're viewing a single thread.
In what universe are polls ever anywhere near within half a point?
2018
Sorry this just isn't so. You can't just observe a range of polls which predict a wide range of outcomes and probabilities after the fact and pick whichever one turned out to be best after the fact and say polling was accurate unless that poll you picked consistently achieves that level of accuracy.
This would be like rolling a box full of 100 sided dice before a game and then picking the 2 closest to the football games final score and then declaring that those dice are accurate to within so many points.
Polling is not nor has it been accurate to half a point. The pollsters themselves claim a larger margin of error.
This wasn’t meant to be a prediction and moreso a hypothetical because of how much discussion of polling error I hear from both sides. Like, what if the Polls are actually really good and accurate this time? This is what we’d get
You do realize that voting trends in non-presidential election years are markedly different than those in presidential election years, right? Not just turnout, but actual voting patterns?
Hence this is meant to be a hypothetical. As I said, if it's a 2022 repeat Harris sweeps the swing states and if it's a 2020 repeat Trump sweeps them.
And that leads into my point - the polls from year to year are never “broken” in the exact same way, especially in the 21st century. It’s not going to be a repeat of 2016, 2018, 2020, 2022 - to assume so, no matter the projected outcome, is an inherently flawed hypothesis out of the gate.
- We’re still in a Terra Incognito of post-pandemic polling and voting behavior.
- We have a candidate who is a convicted felon, and who was involved in an attempt to subvert our democracy after the last election.
- We have another candidate who won no primaries to get their party’s nomination, but who quickly locked up party support after a sitting President abandoned their run for reelection. Oh, and she’s a woman with immigrant parents of differing ethnic backgrounds.
- Unprecedented economic situation where all the economic indicators are showing one reality but a lot of people’s lived experience is exactly contrary.
- Huge campaigns to influence the outcome involving entrenched power structures, foreign governments, and ideological extremists, all made easy by AI and tech bros who have abandoned democracy.
The list goes on. Nothing about this election is normal, and almost all of it is unprecedented. The disparity of the polls from the results will of course exist, but the why of that disparity will greatly influence in what way the numbers are off. Did black non-college educated women show up for Harris. Did white suburban soccer moms still embrace Trump. So it’s not simply a question of “the polls are going to be off by a predictable amount, and so we can guess as to which way the election will turn based off past performance”
This wasn't meant to be a prediction and moreso a hypothetical because of how much discussion of polling error I hear from both sides. Like, what if the Polls are actually really good and accurate this time? This is what we'd get
That maybe could happen, but I say it’s much, much more likely they’re flying blind. Polling has become exponentially more difficult over time and there’s only so much you can do to try and address that with your methodology.
Ironically we’re pretty good at identifying trends in how sentiment is changing or shifting, but absolutely crap at being usefully predictive in terms of results.