We are four and a half months out from Election Day 2026, which means forecast season is officially beginning.
On today’s episode of the GD POLITICS podcast, I spoke with two election forecasters whose models are beginning to shape how we understand the midterms: Lakshya Jain, head of political data at The Argument and CEO of Split Ticket, and Zachary Donnini, head of data science at VoteHub.
VoteHub’s midterm forecast is officially live, but Split Ticket’s is set to publish later this week, meaning listeners are getting an exclusive preview of the work Lakshya has been doing.
The forecasts agree on the big picture: Democrats are favored to win the House, while the Senate is close to a toss-up, with Republicans holding a slight edge. But under the hood, the models diverge in meaningful ways.
VoteHub gives Democrats a 72 percent chance of winning the House, while The Argument/Split Ticket puts the odds at 90 percent. The biggest reason is how the two models treat the national political environment. VoteHub’s model is built around a Democratic advantage of about seven points on the generic ballot. The Argument/Split Ticket model, relying in part on its own likely-voter polling, sees something closer to a nine-point Democratic environment.
The two also differ on whether to incorporate prediction markets. VoteHub does, though Zach emphasized that Kalshi markets are weighted lightly in low-volume House races and more heavily in higher-volume Senate races. His argument is that prediction markets can sometimes pick up information before polls do, especially from late-breaking scandals or meaningful early-vote data.
Lakshya is more skeptical. He sees value in prediction markets, but worries about feedback loops and overreactions.
The Senate picture is even more interesting. The two forecasts are almost identical at the chamber level: VoteHub gives Republicans a 55 percent chance of retaining the Senate, while The Argument/Split Ticket puts it at 53 percent.
But the race-level forecasts differ substantially. In Georgia, VoteHub gives Sen. Jon Ossoff an 87 percent chance of winning reelection. The Argument/Split Ticket puts him at 98 percent. Lakshya argues that Georgia is simply not red enough, especially in a Democratic-leaning national environment, to justify treating Ossoff as vulnerable. Zach agrees Ossoff is favored, but his model is more cautious because it adjusts for the possibility that competitive states in 2024 were artificially bluer than their underlying partisanship.
Other divergences tell us a lot about how forecasting works. In Michigan, Lakshya said he thinks VoteHub’s forecast is probably better than his own, because early polling suggests Democrats may not be as strong there as his fundamentals-heavy model currently implies. In Florida, Zach said Lakshya’s model may be capturing something VoteHub is not: the continued Republican strength among Florida’s older and Hispanic voters.
That was the spirit of the whole conversation. Lakshya and Zach are not dueling forecasters, but friends with different ideas about what inputs to use in a model. And that may be the most useful takeaway. Forecasts are not magic. They are structured arguments about what matters, what data should count and how uncertain we should be.
In 2026, those arguments point to a Democratic edge in the House, a highly competitive Senate and a midterm that could determine whether Trump spends the final two years of his presidency constrained by Congress or empowered by it.
