Update Log for the Georgia Senate Runoff

Looking for the Georgia Runoff Forecast? Click here!

December 6th, 2020

New Page for Polls - Now, I’ve made it easy to keep up to date with the latest polls which are under the same menu tab as the Runoff. You can access it here.

December 4th, 2020

RMG Research Poll Split Into Two Parts - RMG Research released a poll today that included voters that say they will definitely vote, are very likely to vote, and are somewhat likely to vote. We don’t know what the turnout will look like, so I broke this into two polls . The first limits the poll only to those voters that say they will definitely vote, which favors Perdue and Loeffler. The other includes the larger sample, which favors Ossoff and Warnock. Each part gets half credit, so RMG Research doesn’t get any extra weight in the forecast.

December 2nd, 2020

Adjusted to the Simulation - Every update, I run 50-100k simulations of the forecast, and each simulation uses the projected margin for both Georgia races, and randomly has both candidates perform or underperform their projections within a reasonable range based on how much uncertainty there is about the outcome of the Senate race.
Obviously, neither of these races are happening in a vacuum - and I combine a great deal of the randomness so that both Democrats or both Republics will likely outperform their projections. Previously, 75% of their randomness was correlated, and today, I raised it to 85%.
I think turnout will be the determining factor in this election, and I made the adjustment so the model will more accurately reflect this. As of today, the chance of a split decision will drop to a lower (and I believe more accurate) number, raising the chances of a Democratic or Republican outright win.
Of course, a Warnock-Perdue victory is still a viable outcome, but it's less likely than the other alternatives.
2. Added New Trend Line For Win Percentage - in line with this change, I thought it was particularly important to show how the chances of a split and Democrats/Republicans winning both races have changed over time. I just added that to the interactive graph showing win percentage over time. You can still see how the Win Percentage trend has changed for the individual races by clicking the left or right arrow next to the box that says “Runoff - All”.

November 30th, 2020

The Warnock Penalty has been removed, now that the 20 day period has ended. Warnock and Perdue are still a modest favorite in their respective races. The most likely outcomes are that Democrats or Republicans win both races, but a split victory, most likely with Warnock and Perdue becoming Senators, remains viable.

Warnock had the highest favorable of anyone running on election day, while Loeffler remains highly unpopular. She’s likely going to remain unpopular because every single ad she is running is negative. She’s running a hardline campaign and has even reached into the ghost of the 2008 election by bringing back Reverend Wright to attack Raphael Warnock. The Reverend Wright attacks on Obama are a textbook example of racial dog-whistling (the kind McCain decided not to run). She clearly believes her pathway to victory is entirely through base turnout and not through persuasion, which is why I suspect she will finish with a lower percentage of the vote than Perdue.

November 10th, 2020

1. Warnock Penalty

I’m temporarily giving Warnock a modest 2.0% penalty, that will fade away by 0.1% a day until it completely disappears in 20 days. Warnock has a remarkably high near 20% approval rating and a robust lead in the polls. While his skillset as a candidate has undoubtedly contributed to this position of strength, it’s also likely a result of a near absence of negative ads against him, as Kelly Loeffler and Doug Collins focused their fire on each other. Chance is, he will become a bit less popular and the race will get more competitive over the next few weeks. I could have simply waited for this to happen - but my goal is to have the forecast predict as accurately as I can even this far out.

2. Fundraising:

I've significantly reduced the impact that fundraising has on the Forecast. Normally, the number of money candidates raise from individual donors relative to each other is an excellent indicator for predicting the election. This was absolutely not the case in 2020 - and I think it led Race to the WH to overestimate Democrat's performances. I anticipate this is a result of historic levels of enthusiasm among the hardcore Democratic base, primarily because of their extreme dislike for President Trump. This broke this component of the model.
Normally, fundraising is so predictive not only because it shows which candidates have more resources, but because it signals that a candidate has a strong message and has the talent to activate grassroots support. However, almost every even slightly competitive Democrat candidate raised incredible sums of money. With the fate of the Senate resting on Georgia, fundraising is unlikely to be a strong determinant of skill set this time around either. Therefore, its weight has been reduced to 35% of its levels in the normal 2020 Forecast.