Democrats are favoured to win control of both chambers of the US Congress in the 2026 midterm elections, according to prediction markets and polling-based models. However, prediction markets are slightly less confident about the party's chances than poll-based forecasts.
Prediction markets give Democrats a 94% chance of winning control of the House of Representatives, compared with a 96.2% chance estimated by polling-based models, according to Yahoo Finance's analysis on Tuesday.
Democrats have a 77.3% chance of taking control of the Senate, according to prediction markets. Polling-based models put their chances higher at 80.2%. The difference between the two estimates has narrowed as Election Day approaches.
Deep Dive
How do prediction-market forecasts for the 2026 midterms differ from polling-based models?
How are prediction-market odds for the 2026 midterms calculated?
Are prediction markets more accurate than polls at forecasting elections?
The difference between prediction markets and polling models has been more noticeable in the Senate race. According to Yahoo Finance's analysis, the gap between their estimates has averaged around 7.3 percentage points since September 1.
US midterms 2026: Democrats lead in House and Senate forecasts
However, this gap narrowed to just 2.9 percentage points on Tuesday morning, showing that prediction markets and poll-based forecasts are moving closer to each other as voting approaches.
{{/usCountry}}However, this gap narrowed to just 2.9 percentage points on Tuesday morning, showing that prediction markets and poll-based forecasts are moving closer to each other as voting approaches.
{{/usCountry}}Also read: US midterm results 2026: When will winners be declared? All you need to know about vote counting
The difference between the two methods has been much smaller in the House race. Their average gap has remained below one percentage point over recent months, although Tuesday's estimates showed a difference of 2.2 percentage points.
Yahoo Finance calculated these estimates by combining data from three major prediction markets: Kalshi, Polymarket and PredictIt. It averaged their daily odds to create a single prediction market estimate. These figures were then compared with two polling-based models. These included a model developed by statistician Nate Silver and another called ElectIt, developed by Harvard.
The difference between prediction markets and opinion polls has become an important part of the debate over how accurately US elections can be predicted. This debate gained attention after the 2024 presidential election, when prediction markets gave Donald Trump better chances of winning than many polling forecasts.
Prediction markets and opinion polls had shown a difference of around 10 percentage points before the 2024 election. This led to several post-election analyses claiming that prediction markets had correctly predicted the outcome while traditional polls had failed.
Prediction market companies such as Kalshi and Polymarket have repeatedly argued that their forecasts can be more accurate than traditional polls because people trading on these platforms have money at stake.
However, prediction markets have also faced forecasting failures. One recent example was Brazil's presidential election, where the eventual winner had been given only a 20% chance of victory by prediction markets before winning.
The prediction market industry is also seeing a sharp increase in trading activity ahead of the 2026 midterms. Kalshi shared data over the weekend showing that more than $1.2 billion had been traded on its midterm election markets. This trading volume has already surpassed the platform's total volume during the 2024 US presidential election, highlighting the growing interest in betting on political outcomes.
Government scrutiny of prediction markets increases
The rapid growth of prediction markets has also attracted greater attention from US lawmakers. Policymakers have raised concerns about election-related trading, possible market manipulation and the influence these platforms could have on public opinion.
Democratic Senator Elizabeth Warren and Republican Senator John Curtis recently joined hands to raise concerns about election-related prediction markets. They sent a letter to Kalshi and Polymarket questioning the risks linked to these markets.
Warren and Curtis said they were particularly concerned about prediction market contracts linked to elections. They pointed out that these markets receive significant media coverage and are sometimes treated as indicators of political momentum, according to Yahoo Finance.
Also read: US stock futures mixed as Apple slides 4.5%, investors await midterm election results
Warren could lead the Senate Banking Committee if Democrats win control of the Senate in the midterms. This could give her an important role in overseeing the financial sector, including prediction market companies.
Prediction markets are also facing legal challenges over whether federal regulations can override state gambling laws. In Utah, Kalshi has filed a lawsuit challenging the state's efforts to apply its anti-gambling laws to prediction markets.
A group called Better Markets has supported Utah's position through a legal brief. The group argued that states should have the right to protect their residents from the harmful effects of gambling, including sports betting.
What happened in the 2024 US presidential election?
The current difference between prediction markets and polling models has brought back memories of the 2024 US presidential election, when both methods gave different estimates of the candidates' chances.
Prediction markets gave Donald Trump a higher chance of winning the 2024 presidential election than several traditional polling forecasts. Trump eventually defeated Democratic candidate Kamala Harris.
On the day before the election, prediction markets gave Trump between a 57% and 62% chance of winning. However, many opinion polls showed the contest between Trump and Harris as almost evenly matched, as noted by Yahoo Finance.
On election morning in 2024, statistician Nate Silver's model gave Harris a 50.015% chance of winning. This estimate was based on an average of dozens of opinion polls. However, the difference between a 50% and 60% probability does not mean that one candidate is certain to win. A 50% chance means a candidate would be expected to win five out of 10 similar elections, while a 60% chance means six wins out of 10.
Prediction markets underestimated Trump's popular vote win
Prediction markets also underestimated Trump's chances of winning the popular vote in 2024. On election morning, Kalshi and Polymarket gave him only a 20% to 30% chance of winning the popular vote. Trump eventually won it.
Despite this miss, prediction market executives celebrated Trump's victory as evidence that their platforms had correctly forecast the election. Polymarket CEO Shayne Coplan argued that the higher odds given to Trump before voting, along with movements in the markets on election night, showed that the platform had successfully called the election.