This is a special mid-week update from The Forecast!

The Model has been updated to Version 3.1.

As I’m sure you’ve noticed, the news for Democrats has been quite bleak the past few weeks. As I dug into the data more deeply to better understand why, I found a formula in the Model that is intended to gauge public sentiment compared to previous elections was not working as intended - it was averaging a column improperly and skewed the results of all previous versions of the Model. This formula has been fixed, and historical tests were rerun to ensure accuracy (you can see those results here).

Below are the updates for the CURRENT election cycle based on these fixes.

The Senate rating has shifted from Solid R → Likely R.

Listed below are the rating shifts:

Shifts toward the Democrats
Colorado (D-Hickenlooper): Likely D → Safe D
Iowa (R-OPEN): Solid R → Likely R
Michigan (D-OPEN): Likely D → Solid D*
Montana (R-Daines): Lean R → Toss-Up*
Texas (R-Cornyn): Solid R → Likely R

*These races have no current polling data, so historical data is informing these predictions. I PERSONALLY would anticipate Michigan to be closer to Toss-Up and Montana to be closer to Likely or Solid R.

You can see how the change in the formula has impacted these races - it is better reflecting a Democratic sentiment that, while still not showing signs of a “blue wave”, is indicative of a trend that leans towards the Democrats.

The House rating has shifted from Lean D → Solid D.

In total, 68 races saw their ratings shift. For the Democrats, they now sit at 220 seats (24 of which are Lean D and are competitive). The Republicans sit at 197 seats (29 of which are Lean R and are competitive). There are now 18 Toss-Up seats.

The shifts, similar to the Senate, were broadly in favor of the Democrats, which is more reflective of the national political sentiment captured in Generic Ballot polls.

The Gubernatorial Forecast actually held steady, still rating both races as Lean D. However, the anticipated margins have improved a bit for Democrats:

New Jersey: D+6.56%
Virginia: D+4.60%

With this fix in place, the Model did see a marginal improvement in accuracy in historical tests. The races for governor in Virginia and New Jersey this year will be the first live test for this new version!

Thank you for bearing with me as I’ve continued to iterate. I’m sorry for the error, and I’m working to make sure things are working properly to provide accurate predictions!

Thanks for reading!