Lede
Early vote data for four different background states now shows a consistent pattern of abnormally high levels of GOP defections to Harris and/or high rates of independents breaking for Harris.
This diary is not about explaining why this phenomenon is happening (e.g., country-over-party movement, the post-Dobbs factor, the Cheney/NatSec alliance, campaign strategy, enthusiasm, etc.), but simply exploring what some of the best available data shows us is already happening.
Background and Method
2024 in many ways is shaping up to be different than other voting cycles. The zone is flooded far more than usual with garbage (that word again) polls by red pollsters. But some 60 million people have already voted. Is it possible to use a combination of high quality, publicly available data sources to find the signal amidst the noise?
The following analysis combines individual voter modeling (courtesy of TargetSmart/NBC) and crosstab polling results for those who have already voted (courtesy of CNN and Marist). Actual voter data is preferable to preconceived models and polls of likely voters, registered voters, or historical active voters. Individual, granular voter modeling substitutes for the weighting step in pre-election polling.
Those checking the TargetSmart/NBC party ID modeling and gender gaps on a daily basis have likely noticed that both are proving quite stable from day to day, even while large batches of daily vote data are added. This lack of volatility does not rule out unpredictability or future anomalies on or before November 5, but it does point to increasing reliability and clarity.
Building on my previous posts about emerging GOP defection rates in Michigan, Wisconsin, and Pennsylvania, and responding to user feedback to simplify the calculations, the following consolidates the scenarios and estimates in one post. Each state has a tailored set of five scenarios for how independent/unaffiliated voters are breaking for Harris, along with the corresponding rates at which GOP-affiliated voters are breaking for Harris.
I've refined the method to be a bit more cautious, with an initial assumption in all states that Democrats are voting 94% for Harris, with 6% leakage going to Trump and others.
Michigan
- Actual early voters: 61% Harris, 35% Trump, 4% others (source: CNN MI, sampled Oct 23-28)
- Early voter party ID: Dem 47%, GOP 42%, Ind 11% (source: NBC/TargetSmart MI, 2.4M voted)
Table: GOP Defection Rates in Michigan's Early Vote
|
Ind -> Harris
|
GOP -> Harris
|
|
80%
|
19.1%
|
|
75%
|
20.4%
|
|
70%
|
21.7%
|
|
65%
|
23.0%
|
|
60%
|
24.3%
|
Wisconsin
- Actual early voters: 60% Harris, 38% Trump, 2% others (source: CNN WI, sampled Oct 23-28)
- Early voter party ID: Dem 35%, GOP 24%, Ind 41% (source: NBC/TargetSmart WI, 1.1M voted)
Table: GOP Defection Rates in Wisconsin's Early Vote
|
Ind -> Harris
|
GOP -> Harris
|
|
65%
|
1.9%
|
|
62%
|
7.0%
|
|
59%
|
12.1%
|
|
56%
|
17.3%
|
|
53%
|
22.4%
|
Georgia
- Actual early voters: 55% Harris, 45% Trump, 0% others (source: Marist GA, sampled Oct 17-22)
- Early voter party ID: Dem: 45%, GOP 48%, Ind 7% (source: NBC/TargetSmart GA, 3.5M voted)
Table: GOP Defection Rates in Georgia's Early Vote
|
Ind -> Harris
|
GOP -> Harris
|
|
80%
|
14.8%
|
|
75%
|
15.5%
|
|
70%
|
16.3%
|
|
65%
|
17.0%
|
|
60%
|
17.7%
|
Arizona
- Actual early voters: 56% Harris, 44% Trump, 0% others (source: Marist AZ, sampled Oct 17-22)
- Early voter party ID: Dem: 34%, GOP 42%, Ind 24% (source: NBC/TargetSmart AZt, 1.7M voted)
Table: GOP Defection Rates in Arizona’s Early Vote
|
Ind -> Harris
|
GOP -> Harris
|
|
80%
|
11.5%
|
|
75%
|
14.4%
|
|
70%
|
17.2%
|
|
65%
|
20.1%
|
|
60%
|
23.0%
|
Corrections, Calculations, Caveats
Corrections
I've now moved all calculations to Excel and stopped using ChatGPT to run calculations, because I found errors in the previous Pennsylvania calculations that I asked ChatGPT to run for me. Pennsylvania is *not* showing patterns consistent with these other battleground states and may be sui generis in this election.
Calculations Explained
The GOP defection rates are calculated from the following:
1. Key Variables:
- H_Early is Percentage of early voters who have already voted for Harris.
- P_Dems is Percentage of early voters registered or modeled as Democrats.
- H_Dems is Percentage of Democrats voting for Harris (assumed constant of 94%).
- P_Inds is Percentage of early voters registered or modeled as Independents.
- H_Inds is Percentage of Independents voting for Harris (varies by scenario).
- P_Reps is Percentage of early voters registered or modeled as Republicans.
2. Formula:
H_Early = (P_Dems × H_Dems) + (P_Inds × H_Inds) + (P_Reps × x)
which, rearranged, is
x = ((H_Early - (P_Dems × H_Dems) - (P_Inds × H_Inds)) / P_Reps
In Excel, it looks like this:
Final Note on Perspective and Motivation
Contrary to one or more comments on my previous posts, these statistical analyses are not an invitation to complacency and certainly not a guarantee of victory. Instead, they provide added impetus to reach out to *all* potential voters that we can. Lots of Republicans and independents can be persuaded, and lots of infrequent and new voters can be encouraged to turn out. Let's run through the finish line!
