With the primary season winding down and the mid-term general election campaign heating up, it’s time to unveil the 2022 Decision Desk HQ Forecast Model
Our model takes into account over 100 variables to predict the probability of the winner for each Senate and House seat. Things like state/district history, polling, incumbency, fundraising and much more (a full list is available on our methodology page) and run millions of simulations to create a win probability for both parties and a mean seat projection in both houses of Congress.
Today we’re making public our first model run and set of predictions.
The top line numbers:
At this point, control of the Senate is a tossup with a slight edge at 54% for the Democrats maintaining control. Our mean seat projection is a continued 50-50 seat split, with Vice President Kamala Harris continuing to serve as the tiebreaking vote in favor of the Democrats.
We rate three seats, Georgia, Nevada, and Pennsylvania as tossups.
You can see our complete Senate forecast, including a seat-by-seat rundown here.
In the House it’s a very different story. According to our model Republicans are the overwhelming favorites at 88% to come out of November with a majority. We have a mean seat projection of 235 Republicans and 200 Democrats.
There are currently 13 tossup seats in the House….9 in Democratic or Democratic leaning districts, 3 in GOP or GOP leaning districts, and 1 new district.
You can see our full House projection here.
Just to give you a sense of how our previous models turned out….
In 2020 our congressional model correctly predicted the outcome of 30 of 33 (91%) US Senate seats and 415 of 435 House seats (95%). On the presidential level our model was the closest to accurately predicting Joe Biden’s electoral vote count.
Our model was featured in the Harvard Data Science Review.
In 2018, the DDHQ forecast model predicted the composition of the Senate within one seat and the composition of the House within two seats.
We expect that the model will change as the data changes during the campaign. That doesn’t mean it’s wrong, just that we’re learning about the campaign and some of the factors driving the election.
As for why we do it, well, we know experts in politics and other fields use these kinds of models to drive their decision making. We think enabling more people to have access to this kind of information is a good thing and we’re happy to be a part of hopefully improving people’s understanding of the election system and helping them to form more informed opinions.
We hope you enjoy digging into our site and look forward to seeing how we do.