Can you forecast an election from voters who never answer a poll? In Ep. 135 of The Grant Williams Podcast, released the day before the 2026 US midterms, Marvin Barth of Thematic Markets argues that the misses of recent cycles were not random error but bias, produced by low-trust populist voters who skip surveys yet turn out when a populist is “on the menu”. He concedes that none of his evidence is statistically significant and makes the call anyway, as a judgment . Then he follows it to its consequences: a Donald Trump who, if the Republicans hold Congress, would be at his most powerful in his final two years rather than a lame duck . Everything here is a forecast, offered before any result was known.
Key Topics
- Barth sets out the consensus of a Democratic sweep, with the generic ballot (the national poll that asks which party, not which candidate) at seven to nine points for the Democrats and betting markets at “92 to 95 percent” for a Democratic House .
- He recalls calling 2016 against the polls and explains it by populist voters who, in his view, “tend to be low trust voters” and “don’t answer polls” , a bias he traces back to 2014 .
- He defines populism not as a demand for “free stuff” but as political disenfranchisement, drawing on his 2016 study “The Politics of Rage” , and names a surge in left-wing populism as a risk to his own analysis .
- Barth calls his method Bayesian in shape, starting from a “prior”, the view one enters with , and argues that reweighting can make the bias worse when only high-trust people answer , citing David Shor’s finding that trust, not class, divided Trump and Clinton voters .
- His call: the Republicans hold the Senate and may gain a seat , have “a decent chance” in the House , and have a 20 to 30 percent chance of gains in both chambers , something the president’s party has done in a midterm only four times since the Civil War .
- Reading Trump through “what would Jackson do?” and his wish to rank with Jackson, FDR and Lincoln , Barth predicts he would force the Iran war to a conclusion through the blockade and, after Andrew Jackson, the one president who paid off the whole US debt , turn to fiscal consolidation .
- Grant doubts voters will give up “the goodies” ; Barth sketches tax rises on top earners funding cuts at the bottom , tariffs that tax the bottom 90 percent , and entitlement reform under cover of fraud stories .
- They close on “Mamdani rage”, the risk of a hidden left-wing vote , and on betting markets, which Barth says price objective risks well but not uncertainty in the mathematical sense, a risk with “no way of assigning a probability” (what economists call Knightian uncertainty) .
Notable Quotes
"They are reliable voters when a populist is on the menu."Marvin Barth, on low-trust voters, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 9:44
"That to me is what defines populism, not whether you're left or right."Marvin Barth, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 20:46
"And so the only way you ever see this is ex post."Marvin Barth, on polling bias, The Grant Williams Podcast: Forecasts and Fortunes, Episode 1 (2 Nov 2026), 27:48
Outline
- 0:03:36 — The consensus: a Democratic sweep
- 0:08:00 — A record against the grain
- 0:09:23 — Low-trust voters and polling bias
- 0:16:01 — What populism is, and whether it can turn left
- 0:24:43 — Priors and the Bayesian process
- 0:28:35 — Reweighting and its limits
- 0:35:00 — A judgment, not a significance test
- 0:38:17 — The Senate, the House and the call
- 0:49:41 — A president at his most powerful
- 0:52:39 — Two rules for reading Trump
- 0:56:49 — The Iran war and fiscal consolidation
- 01:06:40 — Mamdani rage, betting markets and uncertainty
Mentioned Resources
People
- Marvin Barth (researcher behind Thematic Markets and the episode’s guest)
- Donald Trump (Republican president whose power after the midterms is the forecast’s main stake)
- J.D. Vance (US vice president, the Senate tie-breaker in the consensus case)
- Joni Ernst (Iowa Republican senator who outran her polls in 2014)
- Tom Cotton (Arkansas Republican senator who outran his polls in 2014)
- David Shor (Democratic data scientist and pollster who identified the trust problem)
- Hillary Clinton (2016 Democratic presidential candidate)
- Barack Obama (former president whose 2012 campaign employed David Shor)
- Jean-Luc Mélenchon (leader of the French left)
- Franklin D. Roosevelt (New Deal president whose party gained House seats in 1934)
- George W. Bush (former president whose party gained House seats in 2002)
- Abraham Lincoln (one of the presidents Barth says Trump wants to rank with)
- Andrew Jackson (seventh US president, namesake of Jacksonian populism)
- Thom Tillis (North Carolina Republican senator)
- John Cornyn (Texas Republican senator Barth says his party replaced)
- Howard Lutnick (Trump administration official who floated taxing top earners)
- Kevin Warsh (Trump appointee Barth predicted would prove a hawk)
- Susan Collins (Maine Republican senator)
Organizations
- Thematic Markets (Marvin Barth’s research firm)
- Democratic Party (the consensus favorite in the midterms)
- Republican Party (the party Barth thinks the polls underrate)
- Democratic Socialists of America (socialist movement Barth calls left-wing populist)
- MAGA Movement (Donald Trump’s political movement)
- Cook Political Report (election analysts whose House race ratings Barth uses)
- Die Linke (German left party)
- Polymarket (prediction market)
- US Senate (the upper chamber of Congress)
- House of Representatives (the lower chamber of Congress)
Places
- Maine (New England state with a popular Republican senator)
- Michigan (Rust Belt state with a close Senate race)
- Minnesota (state with a close Senate race)
- North Carolina (state whose Republican Senate seat Barth expects to be lost)
- Texas (state whose Senate race became competitive)
- France (Europe’s fiscal warning case)
- United Kingdom (home of the 2016 Brexit vote)
- Rust Belt (the former industrial Midwest)
- Europe (where left-wing populists may be beating their polls)
Works
- “The Politics of Rage” (Barth’s 2016 study of populist parties in 22 democracies)
- Seriously Marvin (Barth’s free series of shorter pieces)
- One Big Beautiful Bill Act (the 2025 Republican budget law)
- “La Cosa Nostra Americana” (Barth’s framework for Trump’s foreign policy)
- The Grant Williams Podcast (Grant Williams’ interview show)
- Obamacare (the Affordable Care Act)
Concepts
- Populism (in Barth’s definition, political disenfranchisement)
- Low-Trust Voters (populist voters who skip polls but turn out)
- Polling Bias (polling error that runs in one direction)
- Generic Congressional Ballot (the national party-preference poll)
- Bayesian Statistics (updating a prior view as data arrive)
- Statistical Significance (the classical test Barth’s work does not meet)
- Poll Weighting (adjusting samples to match the population)
- Jacksonian Populism (the tradition descended from Andrew Jackson)
- What Would Jackson Do (Barth’s rule of thumb for predicting Trump)
- Left-Wing Populism (populism on the left)
- Gerrymandering (drawing districts to secure a party’s result)
- Lame Duck (a final-term president losing power)
- Coattails (a presidential candidate’s pull down the ballot)
- Fiscal Consolidation (cutting a deficit through spending cuts and tax rises)
- Tariffs (taxes on imports)
- Immigration (Barth’s main example of voters ignored by both parties)
- Neoliberalism (the free-market consensus of the center parties)
- Betting Markets (markets that price the odds of events)
- Knightian Uncertainty (risk that cannot be given a probability)
- Mamdani Rage (the risk of a hidden left-wing vote)
- Midterm Gains for the President’s Party (a midterm rarity)
- Primary Balance (the budget balance excluding interest)
- Independent Voters (voters registered with neither party)
- Entitlement Reform (cutting or restructuring benefit programs)
- Social Security (the federal pension program)
Events
- 2026 US Midterm Elections (the subject of the forecast)
- 2014 US Midterm Elections (Barth’s precedent for large polling bias)
- 2016 US Presidential Election (Trump’s first win, against the polls)
- 2020 US Presidential Election (Trump’s narrow defeat)
- 2024 US Presidential Election (Trump’s return)
- Brexit (the 2016 UK vote to leave the European Union)
- Iran War (the US conflict with Iran)
- COVID-19 Lockdowns (the 2020 lockdowns, Barth’s example of uncertainty)
Closing
Released the day before the vote, the episode puts an out-of-consensus forecast and its reasoning on the record, and makes the case that judgment has to take over where polling statistics cannot reach.