US congressional midterm elections: Prediction markets favor Democratic control of both chambers, but Senate polling remains fragile

Overview

This relaunched monitor assesses the approaching US congressional midterm elections using prediction markets, polling, race ratings, fundraising data, and historical election comparisons. Goldman Sachs sees current conditions as strongly favorable to Democrats in the House and constructive, but less secure, in the Senate.

Core views

Goldman Sachs reintroduces its US Election Monitor to provide periodic updates on the congressional midterms. Its current snapshot is favorable to Democrats: prediction markets imply more than a 60% probability of a Democratic Senate majority and more than a 90% probability that Democrats gain at least the three House seats needed for control. Markets price more than a 60% chance of a Democratic sweep of both chambers and less than a 10% chance that Republicans retain the House. The House outlook rests on the national generic ballot and the distribution of vulnerable districts. Democrats lead the generic ballot by 8.5 percentage points on average, a roughly 11-point swing from the 2.6-point popular-vote margin won by House Republicans in 2024. Democrats' average advantage has widened by 2 percentage points since the start of September to cycle highs. Republicans hold 26 seats in districts President Trump won by less than 11 percentage points; the report identifies these as the districts that markets and analysts generally expect to flip. While mid-decade redistricting has moved several Democratic-held seats into Republican-leaning districts, Goldman Sachs judges that redistricting would likely net the GOP only a few seats if current polling is accurate. Prediction markets imply a greater than 50% chance that Democrats gain 24 seats, well above the three-seat threshold for House control. The report notes that gasoline prices and the Democratic generic ballot have moved together this year. The Democratic lead widened gradually after the gasoline-price spike at the onset of the Iran war, and the two series have recently moved in the same direction. It also highlights issue salience: inflation and the economy are voters' highest-ranked issues, while President Trump's net approval on both is deeply negative. These conditions provide context for the Democrats' current national polling strength. The Senate remains more competitive. Democrats hold 47 seats and need a net gain of four to win a majority. Polling averages show Democrats ahead in six Republican-held seats and in their own close contests, but most leads are within the margin of error. Prediction markets and expert ratings widely expect Democrats to gain North Carolina, while polling shows Democratic leads in five additional Republican-held seats; markets imply Democrats are more likely to win all but one of those contests. Democratic odds have risen after polls in key states showed Democrats leading, with Texas now treated as the marginal race and markets implying roughly 60% odds of a Democratic win there. New Hampshire, currently held by Democrats, has tightened to a Cook Political Report Toss-Up, while Kansas and South Carolina have polled more tightly than expected even though markets still give Republicans a large advantage in both. Goldman Sachs interprets the convergence of overall Senate-majority odds with odds for the marginal Senate seat as evidence that markets see either multiple Democratic paths to a majority, greater correlation across races, or both. However, it stresses that present polling may overstate Democratic prospects. Historical midterm evidence suggests polls tend to understate Democrats in Democratic-leaning states and Republicans in Republican-leaning states, consistent with voters returning to their party late in the cycle. If that pattern persists, current Democratic leads in Republican-leaning Senate races would be weaker than polls imply. The report also finds that Senate polls at a comparable point in the past two midterm cycles have, on balance, overstated Democratic performance. Campaign finance is a further counterweight to polling. Republicans have outspent Democrats in most key Senate races, with Democrats ahead only in Georgia, Kansas, and South Carolina. Republicans began cumulatively outspending Democrats in key Senate contests only in September, meaning current polling may not yet reflect the full impact of that spending. In House races, aggregate candidate spending and cash on hand lean Democratic, but Republicans lead most head-to-head matchups. The report cautions that the available figures exclude estimated advertising over the next two months, which has skewed heavily Republican, and that candidate spending and cash-on-hand data are lagged because disclosures are quarterly. Finally, Goldman Sachs places current prediction-market pricing in historical context. Using Iowa Electronic Markets, PredictIt, Kalshi, and Polymarket data, it finds that the challenging party's odds of sweeping both chambers have on average declined as Election Day approaches. This historical tendency, alongside the polling and spending caveats, qualifies the otherwise Democratic-leaning current market signals.

Analysis framework

The report combines national and state polling averages, prediction-market probabilities, Cook Political Report race ratings, district-level 2024 presidential margins, campaign-finance data, issue-salience surveys, and historical midterm comparisons. It first assesses House and Senate control probabilities, then tests those indications against seat-level competitiveness, spending, and prior-cycle polling and market errors.

Methodology notes

• Other

  Prediction-market probability tracking

  The report uses market-implied probabilities from Kalshi and Polymarket to gauge the likelihood of individual races, chamber control, and a Democratic sweep, while comparing current odds with prior election-cycle patterns.

• Other

  Generic-ballot and race-poll analysis

  National generic-ballot polling is used to estimate the electoral swing from 2024, and state-level polling averages are compared with race competitiveness and historical poll errors.

• Other

  Campaign-finance expenditure analysis

  Federal Election Commission data are used to estimate candidate and independent expenditures, with explicit recognition that quarterly candidate disclosures and unbooked future advertising limit the timeliness of the data.

Key data

• Democratic Senate-majority probabilityMore than 60%Prediction-market implied probability; Democrats need a net gain of four seats.
• Democratic House-majority probabilityMore than 90%Prediction markets imply Democrats gain at least the three seats needed for a House majority.
• Democratic generic-ballot lead8.5ppAn approximately 11pp swing from the 2.6pp Republican popular-vote margin in the 2024 House election.
• Generic-ballot change since early September+2ppDemocrats' average national advantage widened to cycle highs.
• Vulnerable Republican-held House seats26 districtsThese are districts President Trump won by less than 11pp.
• Texas Senate race probabilityRoughly 60%Prediction markets imply approximately a 60% probability of a Democratic win in the marginal race.
• Democratic sweep probabilityMore than 60%Based on averaged Kalshi and Polymarket odds.

Impact & implications

The report's current data portray a Democratic-friendly national environment, especially for House control, but it does not treat this as settled. Senate outcomes depend heavily on close contests where historical polling bias and Republican financial resources could erode the apparent Democratic advantage.

Risks

• Most key Senate polling margins are within the margin of error.
• Historical midterm evidence indicates that polls at this stage have tended to overstate Democratic performance, particularly in Republican-leaning Senate races.
• Republicans' fundraising and spending advantage may not yet be fully reflected in polling.
• House spending data exclude estimated advertising over the next two months, which has skewed heavily Republican, and candidate-finance disclosures are lagged.