Democrats are buoyed by the strong early vote numbers, which show Black voters making up a larger percentage of the electorate than in November and higher early turnout in Democratic congressional districts in the state. Both are positive signs for Jon Ossoff and Raphael Warnock, the two challengers, easing fears that the typical voter dropoff that has plagued Democrats in past years would doom the party’s chances with control of the Senate on the line.
Meanwhile, early-vote turnout has lagged in Republican-held congressional districts, likely leaving GOP Sens. Kelly Loeffler and David Perdue with a larger deficit heading into Election Day than they had to make up on Nov. 3, with early voting concluding this week leading up to the New Year’s holiday.
Politico, December 30, 2020
Early voter record turnout for run-off
As of this morning, 2,566,332 people have voted in the 2020 Georgia Senate run-off election. At this point (i.e., six days out) in the 2020 general election, that number was 3,237,422. Today is the last day of early voting in many counties. No votes will be counted until the polls close at 7:00 PM local time on Tuesday January 5th. We don’t know how folks are voting and we don’t know the share of Democratic voters participating so far (Georgia does not register voters by party), but there does seem to be room to grow in some counties (see map below). However, we do know some things about the folks who have voted early already.
Map Source: US Elections Project
The Black share of the early electorate is 31.3%. For the general election, the share of Black voters was just over 27%. Women make up 55.5% of the early electorate, while 64.3% of early voters are 50 or older. Just over 94% of the early voters voted early in the general election, so while there are not a lot of new voters there are enough to affect the outcome of either run-off. Here’s what we know about these demographic categories voted in the general election:
Trump won 60% of the election day voters, but still lost Georgia by 12,000 votes.
Biden won 54% of women
Trump won ages 45-64 by seven points and 65 and over by 12 points*
88% of Black voters went for Biden
Where is the polling?
Pollsters seem to have vanished since the general election. With no other races happening, one would think there would be a lot of polling. While we don’t know for sure why, there are some reasons that may explain it. In the public and media perception, polling took another hit this year. As we’ve discussed in previous posts, the polls were much more accurate than a lot of people think they were. I don’t want to add to the confusion that newspaper headline writers unintentionally create by suggesting exactitude from polling so below I will discuss the error margins on the few polls we have. But, because this perception of polling inaccuracy persists, perhaps one important reason we have not seen much polling in these races is that media sponsors are reluctant to pay for them. Polling is expensive.
Another reason for pollsters to be reluctant to field surveys in the run-offs is the difficulty in polling run-off elections. This is particularly acute this year as it appears turn-out will be higher than any run-off in Georgia history. This complicates turnout models and, in elections as close as these two appear to be, even the best pollster could end up being way off. This relates back to the perception problems with polling as pollsters may not want to contribute to the poor perception the industry has among the public and media right now.
That said, there has been some polling over the past week in both races. Let’s take a look at it.
A new JMC Analytics poll (B/C) finds both Democrats leading in the run-offs:
Jon Ossoff (D) 50%, David Perdue (R) 43%
Raphael Warnock (D) 53%, Kelly Loeffler (R) 44%
The margin of error in this poll is 4.4 points. This means Ossoff has support that ranges from 45.6 to 54.4 percent while Perdue’s support ranges from 38.6 to 47.4 percent. Seven percent of this sample is undecided. That is a big number in a close race. Let’s say that Perdue really has 41 percent support. If the undecideds break for him** then he will have 48 percent. But, if the reported number (43 percent) is correct then he will have 50 percent. If the polling error is in Perdue’s favor*** then he easily wins the race if the undecideds break for him.
A new Trafalgar Group poll (C-) finds closer races with the Democrats also in the lead:
Ossoff 50%, Perdue 48%
Warnock 50%, Loeffler 49%
The margin of error in this polls is 2.99 percent. This means that Ossoff and Warnock both have support from between 47 and 53 percent of likely voters. However, The Trafalgar Group is a highly disreputable pollster. It claims that it knows better than any other pollster how to survey Republican voters, but it absolutely refuses to be transparent about its methodology. It appears more likely that it just gives GOP candidates a few extra points just because. That might say to you that the Democratic candidates are in good shape since they are both up in Trafalgar’s latest poll. But, since it will not be transparent about its methodology, the margin of error is in these races is suspiciously low,**** and because it has an unlikely low number of undecideds (less than two percent) – I do not think this poll tells us anything useful.
Last week, there were two polls from better rated pollsters. (There have been other polls from newer, unrated pollsters – one of which [Open Model Project] I am very interested in following because of their new approach to methodology – but I want to understand what they are doing before reporting their findings.)
On December 23rd, Insider Advantage (B-) released a poll finding Perdue leading Ossoff by one point (49-48) and Warnock up by two points (49-47). Undecideds make up three percent in the Ossoff race and four percent in the Warnock race. The margin of error is 4.4 points. This means Ossoff’s support ranges from 43.6 to 52.4 percent. Warnock’s support ranges from 44.6 to 53.4 percent.
On December 22nd, SurveyUSA (A) released a poll finding Ossoff up by five points (51-46) and Warnock up by seven points (52-45). SurveyUSA uses non-probablity samples and consequently does not have margins of error, but confidence intervals. The CI for this poll is 3.3 points. For our purposes, this can be substituted for the MOE. This means Ossoff may have support from between 47.7 and 54.3 percent. Warnock’s support ranges from 48.7 to 55.3 percent. Before getting too excited about these numbers, the very first line in the polling memo from SurveyUSA says, “Georgia is a hot mess and no opinion pollster could possibly say what will happen when votes are counted in 2 weeks, 01/05/2021. Any outcome is possible, including victories for the 2 Republican incumbent US Senators.” Even the pollsters are concerned that this poll may be an outlier.
Why are we not hearing about internal polls?
Considering the paucity of public polling right now, one would think we might hear more about internal polling. But, while it may make its way into some fundraising emails (it’s always hard to tell what campaigns are referring to when they say they are winning or losing in order to motivate supporters to give), we are not seeing them being publicly reported. So, the true answer to this question is: I don’t know. My best guess is that none of the campaigns trust their own internal numbers enough to make them public.
That said, there is another indicator that Republicans are more confident than Democrats about the outcomes in these races. There have been reports that GOP Senate Leader Mitch McConnell seems not worried and Democratic Senate Leader Schumer thinks GA is lost (and thus is not courting big donors for fear they will not give next time). There are indications that these reports are accurate. This can be seen in McConnell’s opposition to increasing the “stimulus” payment. Both GA GOP candidates have said they support the increase. This appears to be classic GOP deception. They support the increase knowing that McConnell will not let it come to a vote. McConnell will do whatever it takes to keep his position. He will lose it if the Democrats win both seats. That he does not seem concerned suggests he is not.
Schumer reportedly is not going back to the well with big donors for either race. He is concerned that they will not give next time if these candidates lose after all the money put into these races. Considering how much money failing candidates like Amy McGrath, Sara Gideon, Cal Cunningham, and Jaime Harrison raised, it is a fair concern.
Final Thought: Give money to the organizations doing the canvassing and voter contact.
Senate candidates received record amounts of contributions this year and still lost. Ossoff has raised more money than any non-incumbent Senate candidate ever; Warnock has raised a lot too. There is reason to believe that lack of door knocking hurt down-ballot Democratic candidates in the general election. So, while media advertising is important in a state as big as Georgia it is the canvassing that will make the difference at this point.
It’s time to send money to the organizing groups. You can help support canvassers through Unite-Here, the hotel workers union that led the organizing effort to turn Arizona blue over the past 15 years. Another option is to donate to PRO Georgia, which is a network of advocacy and organizing groups in the Peach State. For a good review of how organizing groups helped turn Georgia blue in the presidential race this year, read this. The organizing groups mentioned in the article include SONG Power, Georgia Latino Alliance for Human Rights (GLAHR), and Mijente which collectively knocked on 150,000 doors in Georgia to get out the vote for Biden. Also mentioned is the New Georgia Project, which registers and engages voters. NGP is doing phone banking for the run-offs.
If you want to donate to the campaigns, please match that with a contribution to an organizing group. You can donate through Fair Fight here and through Flip the Senate here.
Notes
* This is a common problem in comparing different datasets. The reporting on the early vote breaks down ages into different categories than the exit polling did. So, we cannot directly compare how ages 50+ voted or how ages 45+ voted. We have to just accept that difference and understand that the comparison is telling us something about how older folks voted in general.
** At the moment, the best explanation for the presidential race being closer than expected from the polling appears to be that undecideds broke nearly unanimously for Trump at the end of the race. However, there were not enough undecideds to make a difference in key states. This may also be a reason why down-ballot Democrats fared worse than was expected.
*** Almost all competitive Senate races that Republicans won in the general election saw the polling error break towards them. So, if we assume that happens again and undecideds break big for the GOP then both incumbents are in good shape. But, both of those things need to happen to ensure, but not guarantee, a GOP victory. Turnout could still change the calculus.
**** The simple margin of error for a poll with a sample size of 1,022 in a population over ten million should be 3.07 percent. With the undescribed adjustments the pollster is doing to increase Republican support in the estimate (Trafalgar is one of the purveyors of the idea of the Shy Trump Voter), the margin of error should be higher than 3.07 – perhaps a lot higher. We cannot replicate their results because they give us no information about how they adjust their numbers.