***UPDATED 1/4/2021***
The long and exciting election season will conclude on January 5th with the Georgia Runoffs. Two remaining elections – David Perdue (R) vs. Jon Ossoff (D), and Kelly Loeffler (R) vs. Raphael Warnock (D) – will determine control of the Senate. In order to have a majority (determined by the Vice Presidential tie breaker), Democrats need to win *both* Senate seats.
We forecast that Democrats have a 41.2% chance of winning the Senate.
This translates to between a one-in-three and one-in-two chance of taking control.
For the individual races, we predict that Jon Ossoff (D) has a 40.4% chance of victory, and Raphael Warnock (D) has a 42.0% chance of winning his seat.
The outcomes of both these seats are highly correlated – there is an extremely small chance that Democrats win one seat but not the other.
What has changed since the General?
The General Election was a lot better for Republicans than many had forecasted.
This was largely due to systematic polling errors that incorrectly favored Democrats in key states such as Iowa, Wisconsin, and Texas. That being said, Georgia had some of the best polls all cycle (an error of 1.5 points incorrectly favoring the Democrats), and fell within the expected margin of error. Even at the Senate level, the party aggregated polling predictions ended up being quite good.
We’ve made some slight adjustments to our poll averaging methodology and the way we blend polls in with priors in order to account for the general.
The second reason why Democratic performance was overestimated was because they had record-beating fundraising numbers throughout the cycle. While this was important in certain races, it didn’t have as much of an impact as one might expect. The model now has this information and is less bullish on the impacts of fundraising.
While recent FEC reports have been good for Democrats, we don’t expect that to shift things significantly in their favor.
What are the polls saying right now?
Our current poll average has Ossoff + 1.2 and Warnock + 1.7. This is because a slew of recent polls have Democrats up (including one that had Democrats up by a margin of greater than 5 points). If one were to make an adjustment based on the error term during the general, that gets you to Perdue + 0.3 and Warnock + 0.2, which puts things in pure toss-up territory.
It’s worth noting that many of the polls come from pollsters that aren’t particularly well known, and should be viewed somewhat skeptically. We also saw several split-ticket polls, which is unusual given the history of double barreled Senate elections.
Why Are We Doing This?
Polling and modeling took a hit in the general election. While we were generally pleased with our model’s performance, it’s clear there is room for improvement.
The reason we did this initially, to provide the kind of information and tools professionals across many industries use when evaluating elections to a wider audience, is still valid. They will still be using these tools going forward and the public should have similar access.
The challenge is to evaluate, learn and improve the process, and the ultimate output.
Transparency has always been a foundational value for Decision Desk HQ. We enjoy sharing our success and we don’t hide from our challenges.
Conclusion
It’s a close race, but Republicans have the advantage. Vote shares in the General were good for Senate-level Republicans. We’ve seen good polling and financials for Democrats, but the General has made it so that the model weighs those variables less aggressively.