If you’re fully exploring the probabilistic landscape in an uncertain electoral environment, you’re inevitably going to end up with some posts that “didn’t age well.”
asserted
that → explore → posts
For example, if Democrats have a good night next Tuesday and Peggy Flanagan easily wins in Minnesota, our newsletter from earlier this month that seriously contemplated Michele Tafoya’s chances might be one of those.
uncertain
that → have → those
If Democrats have a disappointing night, conversely, this post about the South Carolina Senate race will be a candidate for the “didn’t age well” bucket.
asserted
post → have → bucket
But we haven’t examined South Carolina all that seriously.
asserted
we → examine → Carolina
Nor have we given sufficient attention to a possibility that could give Democrats a winning hand in the Senate, even if they fall short in the Palmetto State: disproportionately high turnout among Black voters.
uncertain
they → give → voters
Minnesota and South Carolina, indeed, were two of the three states that The New York Times and Siena University polled in their latest round of surveys, published early this morning.
asserted
Times → poll → surveys
The third state, Florida, produced the least interesting result, with Republican Ashley Moody ahead by 6 points, pretty much spot-on to FLIPR’s polling average there.
asserted
Moody → produce → average
And that state is actually … Minnesota, where NYT/Siena had Flanagan ahead by 14 points, much greater than her 4.7-point lead in the FLIPR average entering the day.
asserted
Flanagan → have → day
The new poll may reflect Democrats coming home despite Flanagan’s weaknesses as a candidate.1 Indeed, our “simple fundamentals” calculation in Minnesota, based on the generic ballot among likely voters and its partisan lean score, pegs it at D +13.7.
uncertain
calculation → reflect → D
Even so, the poll may be on the optimistic side for Flanagan: “probably too hot for Democrats”, as the Times’s Nate Cohn wrote.
uncertain
Cohn → write → Democrats
(Unlike many other pollsters, the NYT/Siena polls do very little herding, and they’ve never been afraid to publish results that differ from the consensus.)
asserted
that → do → consensus
Still, Minnesota comes up as the tipping-point state 6 percent of the time in our simulations, as much as other states that get more attention.
asserted
that → come → attention
Pushing Minnesota from “slightly weird race that could give Republicans a few backdoor outs” to “state that Flanagan will win unless Democrats are having huge problems elsewhere and are screwed anyway” is potentially worth something.2
Paid subscribers get full access to our Election Night livestream with The Argument, GD Politics, and Slow Boring.
The NYT poll in South Carolina is not an outlier
The Times’s poll in South Carolina drew more attention.
uncertain
poll → push → attention
It showed the race tied 48-48.
asserted
race → show → ?
(Republican Darline Graham was ahead of Democrat Annie Andrews by less than a full point if you go to the decimal place.)
asserted
you → go → place
But this poll is not an outlier relative to the polling average in South Carolina, which had shown Graham up by 1.5 points entering the day.
asserted
which → show → day
That’s partly because the NYT survey wasn’t the only new poll of South Carolina to enter our database overnight.
asserted
survey → ’ → database
A new Citadel Poll has Graham, who was appointed to succeed Lindsey Graham after he died in July, leading by 3 points.
asserted
he → have → points
Although in an interesting twist we’ll return to later, the Citadel had Andrews ahead by 6 points among registered voters.3
In my roundup of Senate races this weekend that compared FLIPR’s forecasts against prediction markets, I didn’t take FLIPR’s side in South Carolina.
asserted
I → return → Carolina
At the time, we gave Andrews a 27 percent chance there, as compared to 20 percent for prediction markets.
asserted
we → give → markets
This is what I wrote about the state:
asserted
I → write → state
Here’s where I feel like we start to get into more speculative territory, even though the difference between FLIPR (27 percent for Annie Andrews) and prediction markets (20 percent) isn’t all that large.
asserted
difference → ’ → Andrews
I suppose I have a reflexive concern about taking the bull case on Democrats in South Carolina given two factors.
asserted
I → suppose → factors
First, the much overhyped Jaime Harrison race, where he eventually lost to Lindsey Graham (RIP) by 10 points.
asserted
he → overhype → points
It has a large Black population, a lot of conservative, evangelical whites, and not many swing voters.
asserted
It → have → whites
Even according to FLIPR, there’s little chance — 4 percent — that South Carolina will be the tipping-point state.
uncertain
Carolina → accord → FLIPR
If polls are systematically overestimating Democrats in red states, Andrews’s chances are basically zip even if Darline Graham, Lindsey’s sister, is a weak candidate.
asserted
Graham → overestimate → states
If the polls are accurate about red-state performance, Democrats could pick up South Carolina if they’re having a fantastic night, but at that point they’ve probably already won Texas and so forth and have more than enough seats in hand.
uncertain
they → pick → hand
Ironically, Andrews’s chances have now improved to 27 percent at Polymarket with the new polling, exactly where FLIPR had it previously.
asserted
FLIPR → improve → it
Although FLIPR’s odds for Andrews have also ticked up slightly, to 31 percent.
asserted
odds → tick → percent
That might actually seem ungenerous to Andrews given how tight the polling is, but FLIPR Classic and FLIPR Deluxe blend polls with “fundamentals” and, in the case of Deluxe, expert ratings.
uncertain
Classic → seem → ratings
One point I’ve emphasized in recent interviews is that polls and fundamentals actually aren’t all that out of whack in states like Ohio, Texas and Iowa.
asserted
polls → emphasize → Ohio
For instance, a roughly 11-point swing from 2024, in line with FLIPR’s current popular vote forecast, would bring a state like Ohio, which Trump carried by 11 points in 2024, basically to parity.
asserted
Trump → bring → parity
But that math doesn’t work in South Carolina, which Trump won by 18 points in 2024.
asserted
Trump → work → 2024
Democrats will need basically everything to go right there: first, a serious outperformance for Andrews based on candidate quality and second, turnout patterns that look much different than South Carolina has experienced in the recent past.
asserted
Carolina → need → past
Why did FLIPR’s odds for Andrews improve, however slightly, even if our polling average didn’t change?
asserted
average → improve → Andrews
It’s because of that disparity between polls and fundamentals.
asserted
It → ’ → polls
In a race with no polling at all, FLIPR has no choice but to go entirely on fundamentals.
asserted
FLIPR → have → fundamentals
The more polling there is, and the more reliable FLIPR considers polls to be, the more weight it puts on polls instead — although it never entirely zeroes out fundamentals.
asserted
it → be → fundamentals
Basically, in possession of more data, FLIPR is a bit more confident now that the polls really do show a close race in South Carolina.
asserted
polls → show → Carolina
…and 4 more, not listed.
The Aporia
