Still a coin toss! But The Economist’s forecast now favors Harris for the first time in weeks.
There has to be some money there for someone who wants to put in the same effort as internal pollsters but share it publicly.
To the extent that there’s any “money” in publicly-shared political polling, it’s in getting your (the polling company’s) name out there, and recognized, so that private companies, and political campaigns, will hire you to do their private polling and market research.
I’ll need one of y’all to let us know how significant this is, but this is apparently a pretty big deal? An Iowa poll that shows her with a 47 - 44 lead in that state.
Per the article :
“The results follow a September Iowa Poll that showed Trump with a 4-point lead over Harris and a June Iowa Poll showing him with an 18-point lead over Democratic President Joe Biden, who was the presumed Democratic nominee at the time.”
I think I’ve made it clear that I don’t put much stock in polls these days, but there seems to be a major swing happening in the past few days. Trump won Iowa by 8 points in 2020.
That poll is a big deal because it’s been historically extremely accurate, even in years when everyone else was way off. If anyone is supposed to get it right its them. I expected Harris to be at -6 or so, and that would have been GOOD news. Edit: this is so good its actually sus lol.
Yeah, that’s a bonkers poll result. In the face of all of the herding going on, Ann Selzer has some cojones for publishing that result. If this poll is the canary in a Harris-landslide coal mine she will cement her place as queen of pollsters. If Trump wins big in Iowa (and elsewhere) she will be open for some major criticism.
To be clear, I am very doubtful that Harris is actually winning Iowa. But if she is even within 4 points there it’s very hard to see how she can lose in the rust belt.
ETA: I should be clear that we should completely be expecting to see random polls like this one much more frequently than we have been seeing. If this were one of the fly-by-night pollsters nobody would really care.
ETA2: Selzer has this cred because she will publish contrary polls and has been right a lot. In 2020 she had Trump+7 when the average was Trump+2 (with one poll actually having Biden ahead). Trump won by 8%.
It’s a huge deal.
Seltzer’s reputation is as Silver stated: someone who has “consistently published seeming ‘outlier’ polls only later to be proven right”
If she is close to being right again then areas similar demographically are very likely also leaning Harris’s way and the systemic error overestimating Trump this time may be … quite large.
I did wonder if the New York Times was making a mistake by making their final national poll of the cycle almost two weeks before the election. Of course, good polling is expensive, and the budget perhaps had to be cut somewhere.
This Iowa poll is very encouraging — full stop. But I wonder if part of the superior Harris reading in Iowa is lack of television advertising there.
In fact, this was the exact Selzer result Nate Silver had touted in the past, when he had written previously about poll herding.
Carl Allen (same guy from my earlier post) teaches that a significant percentage of undecideds introduces uncertainty. In today’s Selzer poll of Iowa, a total of 5% answered “don’t know” or “won’t say”. Third-party voters came in at 4%, so Allen would want to have seen Harris at 48% to feel confident in Harris winning Iowa. However: 48% for Harris is well within the margin of error AND Trump would have to make up three points out of the “don’t know/won’t say” cohort.
Upshot: By Allen-style analysis, Harris is in good shape in Iowa. Probably a narrow favorite. She needs for any margin-of-error miss to be no more than mildly in Trump’s favor. He likely has more ground to make up than she does to get to the finish-line percentage of 48-plus.
A bit more Iowa polling drama:
Emerson College is of course a pollster Silver has criticized.
Good to hear! Can’t wait to see once this poll gets layered into the 538 mix - right now they’re showing him with a 96% chance to take Iowa, but with only a couple of polls from September :
I’m starting to wonder if there isn’t a significant demographic shift in play that the pollsters are missing with all of their convoluted weighting designed to avoid missing Trump supporters.
What if, for example, Harris has weakened marginally with black and hispanic men, but strengthened significantly with white women (of all education levels) relative to Biden 2020? You would expect to see this sort of result in Iowa (one of the whitest, least educated states in the country).
It seems plausible, at least, that by weighting on recalled vote and education status (in addition to race and gender) to correct for the 2016 and 2020 misses, polls are missing a significant erosion in Trump’s support among white women. That would, generally speaking, point to her doing better than Biden in Wisconsin, but I’m not entirely sure what it would mean in MI and PA (probably good news, at the margins).
It would also mean GA would look worse than NC, which is also born out by the pollsters that aren’t weighting so heavily.
Harris is certainly not the favorite to win Iowa, even Seltzer shouldn’t make you ignore all other data, but it is a great sign for other Midwest states, and Harris’ best poll of the cycle. If Times finds similar results in their final midwest polling, I’ll feel relatively confident going into Tuesday.
Nate Silver’s Xitter comments related to the Iowa Selzer poll :
51m
So much for my Saturday night plans. Model update and Model Talk incoming.
·
47m
It is incredibly gutsy to release this poll. It won’t put Harris ahead in our forecast because there was also another Iowa poll out today that was good for Trump. But wouldn’t want to play poker against Ann Selzer.
Selzer hasn’t been off by more than 5 points for an Iowa statewide race in the last 12 years (which is just an incredible track record for polling). For the last two presidential races, she was off by one and two points.
If Harris is even down by 2 in Iowa, she should easily dominate the blue wall states. Up by 3? That could mean landslide.
In case my previous post was too veiled, one of the things that makes Selzer a bit unique these days is that she basically doesn’t weight responses very much at all. What the voters she reaches tells her is what she reports, after very straightforward gender, age, race and geographic weighting (as far as I can tell).
There is obviously a big risk in doing that when you have sub-1% response rates, but she has historically done very well with this approach, picking up trends that others have missed.
Good analysis. We can’t know yet for sure, of course, but I think this is likely.
If the Supreme Court winds up being the reason Trump loses I will consider it adequate revenge for 2000.
And, like clockwork, AtlasIntel drops the following:
ATLAS POLL - SWING STATES
— AtlasIntel (@atlas_intel) November 3, 2024
Trump leads in the swing states, with particularly significant margins in Arizona and Nevada. The race remains tight in the key states of the Rust Belt (MI, WI, and PA). pic.twitter.com/UFStAWretz
ETA: I should add that AtlasIntel is basically the opposite of Selzer in methodology. All online, heavily weighted to make their opt-in approach match some sort of reality. All over the map wrt accuracy (did well in the US in 2020 and I think 2022, but terribly in other recent elections).