Another week, another Forecast! This week, further historical testing has honed in on the Model’s accuracy. This has shifted how comfortable the Model is in projecting a winner. Let’s dive right in with what’s changed.

Only two states have shifted since last week, and both are interesting to me.

Shifts Toward the Democrats
Texas (R-Cornyn): Likely R → Lean R

Shifts Toward the Republicans
Colorado (D-Hickenlooper): Solid D → Likely D

Let’s start with Texas. The state has been a focus for Democrats for the last several elections - it’s a big prize, but also presents a big challenge. In 2024, Democrats felt like it was their big opportunity to pull the Lone Star state home. However, in the presidential race, it was a hard lurch towards the Republicans from what Democrats saw in 2020. In the senate races, they saw very marginal improvement. At this point, one of the biggest factors that will impact how competitive Democrats will be is who Republicans choose as their nominee. Right now, incumbent Republican Senator John Cornyn is facing an intense challenge from Texas Attorney General Ken Paxton. Paxton is trying to challenge Cornyn from the right, and appears to be seeing some level of success (polling shows the two are neck-and-neck). If Cornyn wins his primary, the hill Democrats need to climb is much steeper. However, if Paxton takes the win, Democrats would likely improve their odds, mostly for two reasons: one, Republicans would lose the incumbency advantage; and two, Attorney General Paxton is a divisive figure, even in his own party - he was impeached (and later acquitted) by the Republican-majority Texas legislature. This will be an important primary to watch - Democrats have a rough map in 2026, so they’re looking for ANY opportunity to flip a seat.

In Colorado, I think this is largely a change due to the Model’s confidence changing. The last Republican to win a Senate seat in Colorado was Cory Gardner, who lost re-election to John Hickenlooper (the incumbent Democrat up for re-election in 2026) in 2020. This recency, and the historical competitive nature of Colorado’s Senate seats, give the Model a bit of pause with limited polling. I think when polls start coming in, this race will likely shift back into Solid D territory, but we’ll have to wait and see!

The seat changes listed below were impacted by both changes in the Model’s confidence interval AND changes in polling. Here’s what we’ve got:

Shifts Toward the Democrats
MO-02 (R-Wagner): Likely R → Lean R
NC-05 (R-Foxx): Likely R → Lean R
NC-06 (R-McDowell): Likely R → Lean R
NC-10 (R-Harrigan): Likely R → Lean R
PA-11 (R-Smucker): Solid R → Likely R
SC-02 (R-Wilson): Likely R → Lean R
SC-05 (R-Norman): Solid R → Likely R

Shifts Toward the Republicans
AZ-04 (D-Stanton): Solid D → Likely D
CA-47 (D-Min): Likely D → Lean D
CO-07 (D-Petterson): Solid D → Likely D
FL-09 (D-Soto): Solid D → Likely D
IL-13 (D-Budzinski): Solid D → Likely D
MD-02 (D-Olszewski): Solid D → Likely D
NJ-11 (D-Sherrill): Solid D → Likely D
NM-01 (D-Stansbury): Solid D → Likely D
NY-03 (D-Suozzi): Likely D → Lean D
OR-06 (D-Salinas): Solid D → Likely D
VA-07 (D-Vindman): Likely D → Lean D

A LOT of changes this week, in both directions! Now what you’ll notice is a lot of shifts from “Solid” territory into “Likely” territory. I ran tests on additional historical elections using the same kind of data inputs I’m using today, and compared those results to the ACTUAL results of those elections. This helped hone in on a more accurate confidence interval for the Model.

In order for the Model to rate a race as “Solid”, it has to believe there is a 92% or greater chance of that party winning that seat. That' is an extremely high threshold, and several of the races listed above dropped below that with the change in the confidence interval. However, when adding up the Lean, Likely, and Solid seats for each party, the result is similar to what we’ve seen the past couple of weeks: 219 seats for Democrats, 198 for Republicans, and 18 seats being competitive.

The most competitive seat this week is the same as last week: NJ-07 (R-Kean). The district includes portions of North and Central New Jersey, stretching from just west of Staten Island, NY, to the border with Pennsylvania. In 2022 (just after redistricting in New Jersey concluded), Republican Tom Kean ran against incumbent Democrat Tom Malinowski, who he defeated by 2.8%. In 2024, the now-incumbent Republican Tom Kean defeated Democrat Sue Altman by 5.4%. Also in 2024, the district voted for Donald Trump for President, and the Republican for Senate, both by ~2%. The Model predicts Democrats currently have a 49.77% chance of winning the seat.

Finally, the gubernatorial elections! I know you’re probably wondering, “Hey, where are the 2026 races?” They will be up and ready to go the week AFTER New Jersey and Virginia elect their new governors. Speaking of which, here’s where those races stand:

Shifts Toward the Republicans
Virginia (OPEN): Likely D → Lean D

A significant scandal broke out in Virginia, but NOT in the race for governor - instead, it was in the race for Attorney General. You can read more about that here. It’s too early to see exactly what the impacts could be for other Democrats on the ballot, but recent polling (and the confidence interval for the Model) has prompted this downgrade.

Also, I’ve decided to stop posting expected margins for these (and future) races. In the future, I’ll only include information on rating changes. Predicting EXACTLY where races might fall is tricky business, and I’m more comfortable giving the likelihood of victory, not the margin.

That’s it for the newsletter! Check out the website for more detailed information, and keep an eye out for the next Forecast!