Sep 6, 2026 · 1h 15m
Listen from 37:15
Listen at 37:15
Michigan tests Democrats’ coalition and candidate strategy
Poll Me Twice, Shame On Me
The discussion connects one pivotal Senate contest to broader questions about Trump’s midterm impact, Democratic positioning, and the limits of election forecasting.
- 1Michigan’s Senate race exposes the tension between a favorable national environment and candidate-specific liabilities.
- 2Gaza-related turnout, primary baggage, and candidate quality complicate assumptions about Abdul El-Sayed’s coalition.
- 3Nate Silver’s FLIPR model combines polls and fundamentals while treating uncertainty and correlated errors as central features.
Don't miss
Nate Silver explains how FLIPR accounts for correlated polling errors and simulations, then emphasizes why volatile races remain difficult to forecast.
The brief
Nate Silver, Claire Malone, and Galen Druke join Jon Lovett to examine the 2026 midterms, where Trump’s influence collides with the structural disadvantages facing the president’s party.
Michigan becomes the central test: a potentially favorable Democratic environment meets questions about Abdul El-Sayed’s coalition, Gaza-related turnout, opposition spending, and Mike Rogers’s positioning.
The panel warns against turning one primary or poll movement into a sweeping theory of candidate quality, ideology, or progressive performance in a blue-wave environment.
Silver then opens the forecasting model, showing how polls, fundamentals, fundraising, historical data, expert ratings, and simulations interact rather than produce certainty.
The episode’s broader lesson is that election models are structured judgments under uncertainty, especially when polling errors correlate and volatile races resist precise forecasts.
What was said on this episode
36 statements · 18 positive · 12 negative · 2 mixed · 4 neutral
Trump should end the Iran war quickly to improve Republican midterm prospects.
“He would probably end the war in Iran as quickly as possible.”
Listen at 4:09
Trump should remove remaining tariffs after the Supreme Court decision.
“He would undo whatever tariffs still remain after the Supreme Court decision”
Listen at 4:12
Trump’s approval rises when he reduces media exposure and personal prominence.
“whenever he would chill out and step back from the media and put himself a little bit less front and center, his approval rating would go up”
Listen at 4:26
Reducing Trump’s media prominence will not improve outcomes during his second term.
“that kind of strategy, I think, while it may have worked during Trump 1, won't work during Trump 2.0”
Listen at 4:58
Trump must materially improve voters’ economic perceptions to win support.
“What he really has to do is make people feel materially differently about their lot in life when it comes to inflation and the economy and things like that.”
Listen at 5:04
Trump should reverse his Iran-war decision, hypothetically using a time machine.
“what he should do is collaborate with Elon Musk to, like, build a time machine and undo his decision to go to war in Iran”
Listen at 5:40
Recent midterm political environments generally change little after Labor Day.
“in recent midterms, the environment doesn't change that much between Labor Day and the election”
Listen at 10:56
Republicans will emphasize disliked Democratic positions over the next two months.
“what we're going to see throughout the next 2 months is trying to pivot towards, remember all the things that you don't like about Democrats?”
Listen at 11:50
The they/them advertisement was probably effective against Kamala Harris in 2024.
“the they/them ad was probably a pretty effective strategy against Kamala Harris in 2024”
Listen at 12:59
Recent midterm elections almost always hurt the president’s party.
“these almost always go badly for the president's party”
Listen at 13:20
The current generic ballot favors Democrats by six to seven points.
“The generic ballot right now favors Democrats by 6 to 7 points.”
Listen at 18:40
Abdul El-Sayed is below average and will underperform in Michigan.
“El-Sayed is a below-average candidate. He's going to underperform in Michigan”
Listen at 20:08
A Democrat should win Michigan by approximately seven or eight points this cycle.
“a Democrat should be winning this cycle in Michigan by like 7 or 8 points”
Listen at 20:12
Jocelyn Benson is expected to win by nearly ten percentage points.
“The governor, Jocelyn Benson, is going to win by close to 10 percentage points.”
Listen at 20:18
James Tallarico’s fundraising is exceptionally strong in Texas.
“In Texas, I do think, I mean, Tallarico's fundraising is quite astonishing.”
Listen at 22:03
The Texas race is driven more by Ken Paxton’s weakness than Tallarico’s strength.
“I do think it's more about Ken Paxton being a weak candidate than Tallarico being strong.”
Listen at 22:07
More moderate candidates perform better on average.
“more moderate candidates do do better”
Listen at 22:38
National media overemphasize Israel, Gaza, and Palestine relative to voters’ priorities.
“People in the national media care disproportionately about Israel and Gaza and Palestine relative to if you survey voters”
Listen at 24:34
Arab American voters comprise roughly two percent of Michigan’s voting pool.
“it's 2%, roughly, of the voting pool”
Listen at 24:59
Voter turnout is generally overrated relative to persuasion.
“I'm always someone in general who thinks turnout is a bit overrated relative to persuasion”
Listen at 26:20
Trump’s very favorable rating is approximately twenty-one percent.
“his very favorable number is down to around 21%”
Listen at 27:40
Focusing on Mike Rogers’s Senate actions and Trump alignment benefits Democrats.
“if this is an election about what Mike Rogers will do in the Senate and how he will side with Donald Trump, that is the strongest place Democrats could be”
Listen at 30:15
Mike Rogers would vote with Trump approximately ninety percent of the time.
“he's still going to vote with Trump 90% of the time”
Listen at 31:41
El-Sayed moved from trailing Rogers to leading by three or four points in polling.
“Rogers was ahead by 1 or 2 points in the polling average, and now El-Sayed's up by 3 or 4.”
Listen at 34:54
El-Sayed will probably win Michigan by two or three points.
“he'll probably win by, by 2 or 3 points”
Listen at 43:25
Primary polls, especially down-ballot polls, have an empirical error margin near twenty-five points.
“it's like literally a 25-point margin of error”
Listen at 45:35
Democrats should use positive rhetoric, broad inclusion, and two major bread-and-butter policies.
“we need positive rhetoric, a big tent, and, like, 2 big splashy bread-and-butter policies that help people”
Listen at 50:14
Democratic candidates should take distinctively left and right positions on different issues.
“you want to go left on one thing and authentically left on one thing, but go right on another”
Listen at 52:18
Barack Obama won support from both the most liberal and moderate voters in 2008.
“Obama won both among the most liberal voters and the most moderate voters”
Listen at 54:13
Polling becomes more important as an election approaches.
“polling gets more important as you get closer to the election”
Listen at 1:03:24
The election model heavily weights within-state individual contributions.
“We actually put a big multiplier on within-state contributions”
Listen at 1:03:52
FiveThirtyEight’s 2016 simulation architecture was better than competing models’ architecture.
“our simulation architecture was better”
Listen at 1:07:20
Polling errors are correlated across polls or races.
“errors in the polls are correlated”
Listen at 1:07:24
Candidate quality and incumbency matter less in today’s polarized climate.
“candidate quality, things like incumbency matter much less than they once did”
Listen at 1:10:00
The model does not forecast congressional primaries because errors would reach twenty-five to thirty points.
“We don't forecast congressional primaries because the model would have these margins of errors of 25 or 30 points.”
Listen at 1:11:08
Forecast confidence should decrease when polling is sparse or data conflict.
“If you don't have a lot of polling in a race, if you have contradictory data, you should be less confident about those forecasts.”
Listen at 1:11:23
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Books & mentions
Still Counting
The new Crooked Media podcast extends the episode’s focus on elections, polling, and forecasting.