A very simple projection of the November vote

In my earlier post about the current state of voter registrations, I noted that you could see the county-by-county totals in the contest details for the Senate runoff. What that also means is that if you have current (till now, anyway) voter registration totals, you can do a comparison across the counties of where voter registration totals have gone up the most, and how the vote has shifted in recent elections. In doing so, you can come up with a simple way to project what the 2020 vote might look like.

So, naturally, I did that. Let me walk you through the steps.

First, I used the 2020 runoff results data to get current registration totals per county. I put that into a spreadsheet with county-by-county results from the 2012 and 2016 Presidential elections and the 2018 Senate election to calculate total voter registration changes from each year to 2020. I then sorted by net change since 2012, and grouped the 254 counties into three buckets: Counties that had a net increase of at least 10,000 voters since 2012, counties that had a net increase of less than 10,000 voters since 2012, and counties that have lost voters since 2012. From there, I looked at the top race for each year.

First, here are the 2012 big gain counties. There were 33 of these counties, with a net gain of +2,488,260 registered voters as of July 2020.

Romney  3,270,387   Obama    2,792,800
Romney      53.9%   Obama        46.1%
Romney +  477,587

Trump   3,288,107   Clinton  3,394,436
Trump       49.2%   Clinton      50.8%
Trump  -  106,329

Cruz    3,022,932   Beto     3,585,385
Cruz        45.7%   Beto         54.3%
Cruz   -  562,453

Year  Total voters   Total votes   Turnout
==========================================
2012    10,442,191     6,157,687     59.0%
2016    11,760,590     7,029,306     59.8%
2018    12,403,704     6,662,143     53.7%
2020    12,930,451     

The shift in voting behavior here is obvious. Hillary Clinton did much better in the larger, growing counties in 2016 than Barack Obama had done in 2012, and Beto O’Rourke turbo-charged that pattern. I have made this point before, but it really bears repeating: In these growing counties, Ted Cruz did literally a million votes worse than Mitt Romney did. And please note, these aren’t just the big urban counties – there are only seven such counties, after all – nor are they all Democratic. This list contains such heavily Republican places as Montgomery, Comal, Parker, Smith, Lubbock, Ector, Midland, Randall, Ellis, Rockwall, and Kaufman. The thing to keep in mind is that while Beto still lost by a lot in those counties, he lost by less in them than Hillary Clinton did, and a lot less than Obama did. Beto uniformly received more votes in those counties than Clinton did, and Cruz received fewer than Trump and Romney.

Here’s where we do the projection part. Let’s assume that in 2020 these counties have 59.8% turnout at 2018 partisan percentages, which is to say Biden wins the two-party vote 54.3% to 45.7% for Trump. At 59.8% turnout there would be 7,732,410 voters, which gives us this result:

Trump   3,533,711   Biden    4,198,699
Trump  -  664,988

In other words, Biden gains 100K votes over what Beto did in 2018. If you’re now thinking “but Beto lost by 200K”, hold that thought.

Now let’s look at the 2012 small gain counties, the ones that gained anywhere from eight voters to 9,635 voters from 2012. There are a lot of these, 148 counties in all, but because their gains were modest the total change is +243,093 RVs in 2020. Here’s how those election results looked:

Romney  1,117,383   Obama      415,647
Romney      72.9%   Obama        27.1%
Romney +  701,736

Trump   1,209,121   Clinton    393,004
Trump       75.5%   Clinton      24.5%
Trump  +  816,117

Cruz    1,075,232   Beto       381,010
Cruz        73.8%                26.2%
Cruz   +  694,222

Year  Total voters   Total votes   Turnout
==========================================
2012     2,686,872     1,551,613     57.7%
2016     2,829,110     1,653,858     58.5%
2018     2,884,466     1,466,446     50.8%
2020     2,929,965     

Obviously, very red. Beto carried a grand total of ten of these 148 counties: Starr, Willacy, Reeves, Jim Wells, Zapata, Val Verde, Kleberg, La Salle, Dimmit, and Jim Hogg. This is a lot of rural turf, and as we can see Trump did better here than Romney did, both in terms of percentage and net margin. Ted Cruz was a tiny bit behind Romney on margin, but did slightly better in percentage. The overall decline in turnout held Cruz back.

Once again, we project. Assume 58.5% turnout at 2018 partisan percentages. That gives us 1,714,030 voters for the following result:

Trump   1,264,954   Biden      449,076
Trump  +  815,878

Trump winds up with the same margin as he did in 2016, as the 2018 partisan mix helps Biden not fall farther behind. Trump is now in the lead by about 150K votes.

Finally, the counties that have had a net loss of registered voters since 2012. There were 73 such counties, and a net -17,793 RVs in 2020.

Romney     182,073   Obama      99,677
Romney       64.6%   Obama       35.4%
Romney +    82,396

Trump      187,819   Clinton    90,428
Trump        67.5%   Clinton     32.5%
Trump +     97,391

Cruz       162,389   Beto       79,237
Cruz         67.2%   Beto        32.8%
Cruz +      83,152

Year  Total voters   Total votes   Turnout
==========================================
2012       517,163       284,551     55.0%
2016       511,387       286,062     55.9%
2018       505,087       243,066     48.1%
2020       499,370    

Again, mostly rural and again pretty red. The counties that Beto won were Culberson, Presidio, Jefferson (easily the biggest county in this group; Beto was just over 50% here, as Clinton had been, while Obama was just under 50%), Zavala, Duval, Brooks, and Frio.

Assume 55.9% turnout at 2018 partisan percentages, and for 277,148 voters we get:

Trump      187,587   Biden      91,561
Trump +     96,026

Again, basically what Trump did in 2016. Add it all up, and the result is:

Trump    5,012,802   Biden    4,770,351
Trump       51.24%   Biden       48.76%

That’s actually quite close to the Economist projection for Texas. If you’re now thinking “wait, you walked me through all these numbers to tell me that Trump’s gonna win Texas, why did we bother?”, let me remind you of the assumptions we made in making this projection:

1. Turnout levels would be equal to the 2016 election, while the partisan splits would be the same as 2018. There’s no reason why turnout can’t be higher in 2020 than it was in 2016, and there’s also no reason why the Democratic growth in those top 33 counties can’t continue apace.

2. Implicit in all this is that turnout in each individual county within their given bucket is the same. That’s obviously not how it works in real life, and it’s why GOTV efforts are so critical. If you recall my post about Harris County’s plans to make voting easier this November, County Clerk Chris Hollins suggests we could see up to 1.7 million votes cast here. That’s 360K more voters than there were in 2016, and 500K more than in 2018. It’s over 70% turnout in Harris County at current registration numbers. Had Beto had that level of turnout, at the same partisan percentages, he’d have netted an additional 85K votes in Harris. Obviously, other counties can and will try to boost turnout as well, and Republicans are going to vote in higher numbers, too. My point is, the potential is there for a lot more votes, in particular a lot more Democratic votes, to be cast.

Remember, this is all intended as a very simple projection of the vote. Lots of things that I haven’t taken into account can affect what happens. All this should give you some confidence in the polling results for Texas, and it should remind you of where the work needs to be done, and what the path to victory is.

Related Posts:

• So how did my simple projection work out?
• Beware color-coded county maps
• So what happened in the Latino counties?