With two weeks until the Nov. 3 midterm elections, the numbers favor Democrats. How much, and on what terms, depends on a handful of races where one party's lead is a matter of a few points. Decision Desk HQ, NewsNation's election partner, said on Oct. 21 that its model gives Democrats a 78% chance of winning the House and a 56% chance of taking the Senate.

Both figures have moved their way in the past month. The Senate number, though, means the Senate is much closer to a coin flip than the House is.

What the Model Says

A month earlier, DDHQ gave Democrats a 70% chance in the House, with 231 projected seats. The forecast now projects 236 seats, according to NewsNation's report. In the Senate, Democrats' chance rose from 52% to 56%, with a projected 51 seats.

The most likely single outcome is a Democratic sweep, at 54%. A Democratic House and a Republican Senate comes next at 24%, and a Republican sweep at 21%. DDHQ builds its model from three inputs: structural fundamentals, public polling and prediction markets.

The Senate Map

Republicans hold a 53-47 majority, and Democrats need to gain four seats to reach 51. A 50-50 split would hand control to Republicans, because Vice President JD Vance casts the tie-breaking vote. That is why the contest comes down to a short list.

DDHQ rates four races as toss-ups. In Iowa, Republican Ashley Hinson has a 55% chance against Democrat Josh Turek. In Maine, Democrat Troy Jackson is favored at 55% over Republican Susan Collins. In Ohio's special election, Democrat Sherrod Brown has a 53% chance against Republican Jon Husted. In Texas, Democrat James Talarico has a 55% chance against Republican Ken Paxton.

Michigan and Alaska are rated lean Democratic, while South Carolina and Kansas are lean Republican. In South Carolina, Trump traveled on Wednesday to support Sen. Darline Graham, who was appointed to the seat held by her late brother, Lindsey Graham, NewsNation reported. The model gives her a 70% chance in a state Trump carried by nearly 18 points.

The arithmetic is demanding. Three of the four toss-ups, Texas, Iowa and Ohio, are in states Trump won by double digits in 2024, according to the same report. Democrats must win most of them, hold what they have and also pick up Maine, where Collins has survived tough years before. A single loss can leave the party short.

The model's state-level probabilities are all close to even, so a single upset in either direction can decide control.

The House Looks Easier

In the House, control requires 218 seats, and DDHQ projects Democrats at 236. Among races DDHQ highlights, Democrat James Ager has a 52% chance in North Carolina's 11th District and Paige Cognetti 54% in Pennsylvania's 8th. Republicans lead narrowly in Iowa's 2nd, where Joe Mitchell has 57%, in Ohio's 7th, where Max Miller has 52%, and in two Texas districts, where Monica De La Cruz has 55% in the 15th and Carlos De La Cruz 58% in the 35th.

Even in those, the week-to-week movement was toward Democrats, usually by one to three points, as the model showed.

Why the Odds Moved

DDHQ's report does not attribute the shift to one cause. Plausible drivers include the economy, where the Federal Reserve has raised rates, as Blogerroom covered in the Fed's unanimous 12-0 hike, and a labor market that added only 29,000 jobs in September. NewsNation separately reported that corruption has emerged as a top issue. These are my inferences about context, not findings from the model.

Reading Forecasts With Care

A 56% chance is not a prediction. It means that if the election were held many times, Democrats would win the Senate a little more than half of them. A lead of two or three points in a single race is smaller than the usual polling error. Blogerroom reported how Flávio Bolsonaro beat the polls in Brazil's first round, a reminder that late shifts and miscounted electorates can swamp forecasts.

Models also differ in their inputs. DDHQ's draws on prediction markets alongside polls. Other forecasters weigh the same evidence differently, and I have only reviewed this one, so these figures should not be read as a consensus.

What to Watch Next

Early voting is under way in many states. Turnout in Texas, Iowa and Ohio will show whether Democrats can mobilize voters in places that backed Trump. Any movement in Maine would matter, since a loss there makes the path to 51 harder.

My view: the campaigns' use of their time is itself a signal. Trump's trip to South Carolina, to defend a seat the model gives Republicans a 70% chance of holding, suggests Republicans see the map as less safe than it was a month ago.