Democrats are projected to hold an edge in eight of the ten closely watched races in the 2026 US midterm elections, including five of seven Senate contests and both US House of Representatives races, according to a midterm forecast brief released by Psifos Advisory Group, a division of Newark, New Jersey-based Gateway Group One. The forecast was issued as voting got underway on Tuesday, 3 November 2026.

Key Races and Projected Outcomes

Among the seven Senate races tracked in the brief, Democratic candidates are forecast to lead in Texas, Ohio, North Carolina, Georgia and Michigan. Republicans, by contrast, are projected to hold an advantage in Iowa and Florida. Democrats are also favoured in both monitored US House contests and in the race for Georgia Governor.

The calls range from 'Solid' — used for Georgia's Senate race, which the brief assigns just a 2 per cent chance of an upset — to 'Tilt,' applied to Ohio (48 per cent upset probability) and Iowa, where Republican chances are pegged at 62 per cent. In Iowa's open US Senate seat contest, the model gives Republican Ashley Hinson a projected two-party vote share of 51.3 per cent over Democratic state Representative Josh Turek. In Florida, Republican incumbent Ashley Moody is projected to receive 53.0 per cent of the two-party vote against Democratic challenger Angie Nixon.

The Methodology Behind the Forecast

Rahul Mehta, Managing Partner at the Psifos Advisory Group, outlined what he described as four distinguishing features of the firm's approach. 'It starts from evidence, not opinion. Each seat begins from fundamentals: the last presidential vote adjusted for the national mood, incumbency, primary strength and primary turnout. The weights are learned from past races, not set by judgement. It weighs polls by quality. Larger and newer polls count more, pollsters with a partisan lean count less, and thin polling is used but flagged,' Mehta said.

On the online sentiment component, Mehta noted: 'We judge posts for or against each candidate across 12 platforms, correct for each platform's habitual lean, and compare races with one another. Where the volume is too one-sided to compare fairly, as in Minnesota's 5th (46 posts about one candidate for every one about the other), we leave the conversation out rather than let it distort the forecast.'

Mehta added that five sets of predictions were locked on 2 November, with a 2-point rule set in advance for deciding outcomes, and that the firm would score every reading against the actual results after election day.

Accountability and Stated Limits

Kurus J. Elavia, also a Managing Partner at Psifos, said the methodology draws on 20 years of building and operating digital democracy solutions. He emphasised that online sentiment is treated as evidence to be measured and corrected, not as a proxy for votes. 'We have seen how volume, organised campaigns and audiences outside a district can distort it,' he said.

Elavia was candid about the model's limitations. 'New Jersey's 7th rests on one older poll, and Minnesota's 5th has none. We state these limits in the brief because a forecast serves a client best when it shows where it is strong and where it is thin,' he said, adding that the standard the firm holds itself to is whether it can show clients 'how far to trust a number, not just which way it points.'

Context and What to Watch

The 2026 US midterms are widely viewed as a referendum on the current political environment in Washington, with control of the Senate and select House seats in play. Forecast models from consultancies such as Psifos add to a crowded field of election analytics tools, though most carry explicit uncertainty ranges. Notably, the firm's online-sentiment weighting is self-described as having 'no track record yet,' making the post-election scoring exercise a live test of its predictive validity. Results from the races covered in the brief are expected to begin emerging on the night of 3 November 2026.