I'm saying that the Dems have often had success in mid-terms; there's this narrative that GOPs dominate mid-terms that's just wrong.
I get what you're saying. I'm looking at 2010 and 2014 as recent examples, and to a lesser extent, 2012 and 2016. If we have a Democrat as President, I think the House would be likely to revert back to Republican control in 2022.
That depends on two things: 1. Gerrymandering 2. Turnouts If dems can fix the first to an extent and mobilize voters, 2022 doesn’t have to be disastrous.
This one, not so much... EDIT: I want to be clear. I didn't forecast. I'm not picking a fight. Just looking at the statements with interest. I think you've done really well. It's amazing how the polls again failed to capture Trump enthusiasm.
I think @American Brummie and @Chicago76 have been the most realistic in their forecasts, likely because they are really far in the weeds in this stuff (no offense @charlie15). Here, 76 is referring to Brummie's prediction. Even then, Senate and House are both pretty far off. Polling clearly missed a lot of granular data.
I'm actually more proud of this call than the other one. Democrats are likely to have 225-230 House seats, not 242. Turnout will be north of 150 million (just wait -- already at 141 million!). Democrats won't pick up the Senate seats in Maine or North Carolina, likely dooming their chances of chamber control (I do not think the runoffs will be good for us). And if you read this prediction, as long as you overlook Florida and put Georgia in there, I was right. Biden won't close the gap in North Carolina, we sorta know that already. I'm ready to call the biggest personal misses I had in 2020, and I've grouped them for effect: 1) Iowa, Ohio, Montana, South Carolina, Utah, Maine-02 -- these were states I expected to all be within 5pp, and only Iowa even came close. Ohio in particular is amazing to me. Trump goes to Ohio, insults Goodyear, threatens to send factories overseas, belittles every institution in the state, and they come back to him. Ohioans must be into psychopaths as a fetish. 2) Texas, Florida, North Carolina -- I think the Democratic Party needs to think about how they win back Latinos. I would imagine they should talk to their state parties in Colorado, New Mexico, Arizona, and Nevada. 3) Colorado and Virginia -- I had assumed, incorrectly, that the leftward lurch of Democrats in these states would hit some sort of ceiling. After all, the reason that Republicans won them even 16 years ago is because the states were filled with Republican voters! Alas, I now predict Colorado and Virginia to end up more solidly Democratic than New Jersey, Connecticut, and Oregon within another decade. I also had a one decent prediction: Minnesota.
I called it as less favorable, since you were trying to allay fears of Trump f****ery on Election Day. You had Biden ahead in FL, and maybe I'm reading wrong, but had him ahead in PA day of election, which is technically true, but the vote counting has actually made it a bit more tense. House: Dems lost seats, not +9. (more polling issue than your forecasting) Senate: really was not in play for dems to take control. (same parenthetical) Issues 1) and 2) require us to recalibrate our faith in the electorate actually paying some attention and being guided by reason/logic instead of emotions/identity. 3) seems like a solid demographic shift.
I'm gonna wait on the full results, but I don't think the polls could have done anything in most of the country. The Latino/Hispanic vote, sure. But let's ignore that problem and focus on the parts of the country with low Latino vote. I'll take a state that is almost all the way in and compare some hypotheticals to the 538 final forecast. Okay, Wisconsin. Biden wins this state by 20k votes. First, we'll compare Biden in raw vote totals to his Democratic predecessors: 2016: Clinton has 1,382,536 to Trump's 1,405,284 2012: Obama has 1,620,985 to Romney's 1,407,966 2008: Obama has 1,677,211 to McCain's 1,262,393 2004: Kerry has 1,489,504 to Bush's 1,478,120 2000: Gore has 1,242,987 to Bush's 1,237,279 If you're a Democratic operative in Wisconsin, you look at Trump's crossover appeal in Wisconsin, the GOP's traditional strength, population growth, and Obama's results. You say "okay, we need to get 1,600,000 votes. The Republican turnout machine will be good and get Trump to 1,500,000. We win by 100,000 votes. Biden more than bested that. He got an extra 30k. But Trump got so many more votes! It's absolutely insane just how many. It's even reflected in the 538 forecast. They projected Biden should get 53.7% of the vote. If turnout is 1.6 million plus 1.5 million, that's 3.1 million votes. 1.6 million of 3.1 million is 51.6%. 1,630,000 of 3.1 million is 52.6%. That's pretty darn close to the mark! The most-salient polls in the 538 average have Dem vote share as 52, 53, 57, 52, 48, 51, 51, and 49 percentage points, respectively. That's really close to the mark! So what do I think happened? Trump voters just flat-out lied to people when asked. They either told pollsters they weren't voting, or were voting for Biden or a third-party candidate. And the polls missed by an extra 2-3% because Trump voters lied. The most-salient final polls in Wisconsin that were closest to the mark? FOX News, Atlas Intel, and Marquette Law School. The ones that missed the most? The New York Times, ABC News/WaPo, and CNN. It's pretty obvious, at least to me, which polls Trump voters were least likely to be honest toward.
Excellent post. Could it also be that the second someone said "I'm conducting a poll from NYT/CNN/etc.. that the Trumpers just immediately hung up? That way, these polls that have a name associated with the media entities Trump has villainized just never capture the Trump voter? I guess they could be lying too. However, the polls are only as good as they predict actual results. Blaming the participants is kind of is pointless. It even further brings into question the value of polling.
Could be, but pollsters would have demographic info on non-response rates and would (should?) weight voters who match that demographic (aka, other Trump voters) more heavily to account for it. EDIT: Also, I repped your post, 100% because I'm vain and like the compliment.
The two big issues beyond the Hispanic vote and the respondents flat out lying on the registered vote surveys: figuring out who was likely to vote via weightings due to a) 100+ year highs in turnout and b) conducting the election during the worst infectious disease crisis in 100+ years. I’m not sure that the respondents even knew how likely they were to vote in this election. There was a lot of support for Trump to offset Dem turnout gains that came from somewhere. Some of it was obscured by intentional respondent deception. But some of it very likely was soft support for Trump by voters who originally did not intend to vote.
Yes - it is not like this was an unknown issue Nate Silver discussed it extensively last week. Even as turnout rose, models stayed the same because you assume the electorate remains balanced. But in fact a shit tonne of Trump/GOP leaning voters who did not vote in '16 must have turned out
As did NH. Biden will win by 6-6.5 pts and the UNH poll came in at 8pts. All three of those states + WI are high turnout. WI had much more of a grassroots effort from both parties to engage voters than any of the rest. WI also has more white non-college degree voters than the other 3 IIRC. That’s the biggest hint right there. I think under normal circumstances, the balanced approach probably works as well as anything. But both the greater the prior turnout estimate and the greater the prior margin, the less likely this is to hold in races where both sides are heavily investing. In those cases, it’s probably wise to re-stack the incremental turnout. Turnout in WI is normally 70%. If you’re showing B+8 under a 75% turnout assumption and turnout is actually 80%, it should not be surprising the last 5% breaks heavily for Trump. There’s likely to be much more soft Trump/shrug your shoulders and vote for the incumbent support in the 25% that is excluded in the original 75% turnout assumption. Said differently: you can get to 75% turnout in a state like WI where one side turns out more and the other is business as usual. It is very difficult to assume that to be true when turnout is 80%.
I’m not counting chickens or anything...but the champagne just moved from the outside fridge to the inside one.
So my wife is getting ready to leave for the hospital to go teach clinicals. Asked her if she wanted mimosas before she goes. After saying “very funny” she turned to me and said....”if you pop that ********ing champagne without me they will never find your body” looks like I may have to wait a bit to celebrate
My 8y/o daughter is a huge Hamilton fan. It’s been running in our house like Frozen did several years ago. She just woke up and asked about the election. Told her Biden was ahead in Georgia. She busts out singing: Joe Biden’s gonna be President now! Joe Biden gonna be President....Joe Biden gonna be President....joe Biden’s gonna be President now!