Georgia - Trump v. Harris Voter Registration Index Forecast Model

Topic: 2024 Election Forecast

A lead-in on Voter Registration by Party analysis, from my Voter Registration Index Forecast Model preview of Arizona:

Voter Registration by Party is the most accurate of all indicators in assessing the trajectory of upcoming presidential elections in a county or state.  The logic goes like this: if a young black man registers to vote as a Democrat in Philadelphia in August 2008, who do you think he is voting for?  If a white, non-college factory worker from Johnstown, Pennsylvania, switches from Democrat to Republican in September 2016, who do you think he is voting for? If you chose Barack Obama and Donald Trump, respectively, your thinking is correct.

Critics of this theory like to say, “you don’t have to vote Republican just because you’re a Republican,” and to some extent, that is correct.  It takes a little discernment to understand each county and its intricacies.  Please read my pieces on Westmoreland County, Pennsylvania, and last week’s introduction to the 2024 Voter Registration Index Forecast Model, featuring Pinal County, Arizona.

Georgia, a vital state for Trump’s likelihood of reaching 270 or more electoral votes, does not have voter registration by party.  Such estimates from states not registering voters by party must be inferred from states that do have it.  In future pieces, I will use Iowa to model Wisconsin and Pennsylvania to model Michigan.  It seems logical, given geography, to use North Carolina to measure potential change in Georgia, but I decided against it for two reasons. First, the extrapolation of North Carolina’s voter registration by party trend gives me a Trump margin of roughly 9.4%, which I think is too bullish in that it offers an overly aggressive read on a state that is rapidly reclassifying its party registration but changing the presidential margins at a slower rate, which can happen in a major realignment.  That model gave me realistic outcomes for some major Democrat metros, like Mecklenburg and Wake Counties, but unrealistically tight Democrat margins in some longstanding party strongholds with overwhelmingly black populations, and margins that seem too high for Trump in some of his own strongholds based on the simple fact that the population growth doesn’t seem to be there to justify the raw number in his forecasted margin of victory.  The second reason I passed on using North Carolina is because in 2020, for the first time since 1996, it voted to the right of Georgia, rendering margin swings from 2000 through 2016 moot. It would be nice if we knew all true margins for 2020, but for the sake of this analysis, I will run with the certified numbers.

The pessimistic forecast for Georgia isn’t what we want to see if we want Trump to win, but then again, it carries forward the growth of both parties in a trended format as if the 2020 election will replicate itself.  I don’t see that as likely, especially since mail-in ballot requests and returns are way down by volume in Georgia, and are whiter and presumably less Democrat-heavy than they were four years ago.  If the 2020-redux scenario were to repeat itself, Harris would potentially have an edge roughly twice as large as Joe Biden’s slim 2020 certified margin, which was accomplished through counting and double- or triple-scanning mail ballots in the metro Atlanta area, among other things.  I don’t think her ceiling in Georgia is high, most certainly less than a point, whereas it is Trump who stands to win a surprisingly large margin if he can succeed in unraveling the Democrat coalition in core Atlanta, and dent their black margins throughout the vast hinterlands of the state.

This piece focuses on likely outcomes based on registration analysis by county using Duval County, Florida, and Trump’s own registration to results coefficient there. Soon, in a final piece just before the election, I will issue a formal projection blending all analytical disciplines.

---

Duval County, Florida, Voter Registration by Party

First, here are the voter registration indexes with a given year’s presidential results since 2004 in Duval County, Florida.

Duval County is the sixth most populous county in Florida and is situated near the northeast corner of the state, just a stone’s throw over Nassau County to the Georgia line.  It contains metro Jacksonville and a population sample nearly identical to Georgia’s at 49.4% non-Hispanic white, 28.8% black, and 11.3% Latino, among other smaller groups (Georgia is 51.9%/31.0%/10.5% on those three groups, nearly identical).  Also resembling Georgia, you have a mix of socioeconomic demographics, including suburban whites, urban whites and minorities, and white- and blue-collar workers.  There is a slice of Atlanta, Valdosta, and Swainsboro all within Duval County.

Duval started drifting out of the Democrat “Solid South” voting pattern a little earlier than Georgia did and backed Reagan in 1980 while Georgia stuck with Jimmy Carter, a native Georgian.  Duval also stayed with the GOP presidential incumbent in 1992 when Perot cost Bush the Peach State, but from 1996 through 2016, both backed the GOP nominee, and both flipped in 2020 for Biden, though I don’t believe either would have backed him in a standard race free of 2020 shortcuts and cheats.  Of note, Clinton had a lower share of the vote in 2016 than Obama did in 2012, but the margin was less Republican because Donald Trump nearly mirrored Mitt Romney’s vote total from 2012 and left a lot of votes in the third-party column.

Here is a graphic showing Georgia’s presidential margins compared to Duval’s:

---

Analysis

Since every election dating back to 2004, Georgia has voted to the right of Duval County, most closely aligned in 2004, pivoting sharply left for Barack Obama in his first run, nudging back to the right in Obama’s second run, and regressing leftward in Trump’s first run, likely due to third-party Republicans and modest Democrat inroads.  Both “flipped” in 2020.

How tightly related are their respective swings (change from one election to the next)?  Reference the bottom two rows: