November 2020 Early Voting Day Five: How not to look at the early voting totals

From Twitter on Friday:

.
Biden needs a rout in Harris County (#Houston) to have any chance of winning TX. We’ve had 3 days of record-shattering turnout and by tonight may have 500k votes. But so far, Biden probably isn’t getting #BlueWave he needs. Let us explain.

— Zach Despart (@zachdespart) 6:15 PM – 16 October 2020

You can click over to see the thread, but it’s based on the share of votes that came from precincts won by Hillary Clinton in 2016 versus the share of votes that came from precincts won by Trump in 2016. As of Day Three, the same share of votes from Clinton precincts had been cast in this year’s election, which led to the conclusion that Biden was not outperforming Clinton, at least not yet.

There are several problems with this approach. First and foremost is that “precincts” is too rough a measure to use. Precincts are not uniform in size – there are precincts with upwards of four thousand voters in them, and there are precincts with fewer than one thousand, even those with fewer than one hundred. There are precincts that would have gone well over eighty percent for one candidate or the other, and precincts that were close to fifty-fifty. People move, so over the course of four years a given precinct could be quite different in composition or size. And as we have seen, some people have shifted their voting preferences – college-educated white women, in particular – so one’s vote in 2016 isn’t quite as predictive of one’s vote in 2020 as one might think.

There’s also the fact that the main Democratic strategy is just simply adding to their pile of potential voters, and then turning out as many of them as possible. I’ve said multiple times and in several contexts that there are just simply more Democrats in Harris County than there are Republicans. We saw this illustrated very starkly in 2016, when the total number of Democratic voters increased a lot more from 2012 than the number of Republicans did. See here for my explanation of that. The core of the Democratic voter registration strategy is that most of the folks who had not been voting before were people that were likely to support Democrats, and the focus has been on getting them registered check) and then turning them out. That worked quite well in 2016 and in 2018, and it’s the plan for 2020.

Well, what about the data that we have that suggests most of the voters so far are the old reliables? I will remind you, we haven’t even gotten to the starting line for what would have been the early voting period for this year, and there’s already been a ton of votes cast with that full time span to get everyone else out. There are better ways to estimate what the electorate so far looks like, I’ve talked about this before, and it’s based on using a data model on the vote roster that gives every voter a score of how likely they are to vote D or R, and then sum it all up. There’s some assumptions baked in, and the quality of the data varies a bit from cycle to cycle, mostly because underlying conditions change, but on the whole it’s a reasonable picture. One thing we know is that the first Saturday of the early voting period is a banner day for Democrats, at least in Harris County. That was even true in disaster years like 2010. This year we have two Saturdays, so maybe things will be a bit different – for sure, this Saturday will not be the high-water mark for the week, which is a change from other years – but it is a reminder that different people vote at different times.

In fact, in recent elections, it’s been Democrats who have done better on Election Day than in early voting. Here’s a comparison of the straight ticket vote for the last three high-turnout races:

2012 early - 279,619 R, 259,664 D - 51.9% R, 48.1% D
2012 E day - 124,546 R, 147,327 D - 45.8% R, 54.2% D

2016 early - 308,027 R, 333,477 D - 48.0% R, 52.0% D
2016 E day -  93,636 R, 128,553 D - 42.1% R, 57.9% D

2018 early - 298,644 R, 355,861 D - 45.6% R, 54.4% D
2018 E day - 112,010 R, 159,951 D - 41.2% R, 58.8% D

“Early” combines the mail vote with the early in person vote. I skipped the third party straight-ticket vote for this comparison. Obviously, with a lot more of the vote occurring early, that part of the vote has a much greater effect on the outcome. My point is simply that in past years, the early vote was not necessarily indicative of where everything would end up. Dems in general were trailing in 2012 after early voting, and mostly caught up on Election Day. That was also true for candidates like Ann Harris Bennett in 2016, and Lina Hidalgo in 2018.

I should note that this pattern has also held true for the two most recent lower-turnout races, in 2010 and 2014, which were Republican-dominant years. Dems actually cast more straight ticket votes on Election Day in 2010 than Republicans did, though it wasn’t nearly enough to mitigate the losses they suffered. In 2014, Dems lost all three parts of the vote, but by a smaller margin on Election Day.

The one year where this pattern was broken was 2008 – Dems won the early vote, and Republicans won Election Day, though not by enough for the most part to win countywide. Now to be fair, this year resembles 2008 in a lot of ways – Democratic enthusiasm was through the roof that year, and no one was surprised to see the initial results on Election Night. It will not surprise me if Republicans do better on Election Day than they do in early voting. That doesn’t necessarily mean they’ll make up ground on Election Day – it may just mean they lose the day by less, as was the case for Dems in 2014. Remember, the thesis here is there are more Dems than Republicans. That means that if Republican turnout is pretty good so far – and it does seem to be – then it also means they’re going to run out of voters faster than the Dems. I’d submit that’s what happened in both 2016 and 2018.

All this is for Harris County. I can’t speak for other counties. In a place like Denton County, for example, where Dems made great strides in 2018 but were still outnumbered overall as of that year, it may be that the population growth there plus the level of enthusiasm on the Republican side is enough to not only hold off further Dem advances, but increase the Republican advantage. We won’t know, or at least I won’t know, until we start seeing results. The same strategy of registering more voters, on the same belief that they are on balance more Dem than Republican, holds there. How well it works remains to be seen.

And speaking of Saturday numbers, we now have them. The Day Five daily EV totals are here. You can find the daily totals for 2008 and 2012 (and 2016 as well, but I’ve got a separate link for it) here, for 2016 here, and for 2018 here. I’m just going to give these numbers today, because we’re now at a point where the day-to-day no longer makes sense:

Vote type  Saturday    Total
============================
Mail          8,807   67,255
Drive-thru    7,806   49,970
In person    57,675  468,738
Total        74,288  585,963

Derek Ryan sent out an email that covered the first four days, which you can see here. For the table above, I broke out the drive-through votes from the other in person votes, because why not. I’ve been meaning to ask if they’re tracking dropped off mail ballots separately, I need to follow up on that. Having a Saturday be at two-thirds the level of the Thursday and Friday would be deeply weird in a different year, but this year, who can say? I have no idea what to expect this week or next weekend. Early voting hours today are 12 to 7, instead of the usual 1 to 6, so maybe we’ll get 30-35K. If that’s about right, then with some 620K early and mail votes in the hopper as of Monday, the “normal” start for early voting, we’d need to average 60K votes per day to equal the 1.34 million total turnout from 2016. We’ll see how it goes on Monday and Tuesday, but yeah, I do think that’s within reach. It would mean something like 200K (for the lower end) to 400K (for the high-end Chris Hollins-predicted 1.7 million) voters left for Election Day. Like I said, we’ll see what the next week brings. Have you voted yet?

Related Posts:

• November 2020 Early Voting Day Fourteen: Where will we end up?
• November 2020 Early Voting Day Eighteen: Records were broken
• November 2020 Early Voting Day Sixteen: All through the night