The jungle primary for the Texas 6th special election is just under 2 weeks away, and we have a poll, so everyone is freaking out. The source of the trouble is that the lead Democrat is perilously close to the 2nd Republican, raising fears that the GOP could get two candidates ahead of the lead Democrat, and guarantee a victory before the runoff. This is a theoretical possibility, but not actually a real problem, because that poll should not be taken seriously.
This is a district that is 52% white by population - remember, this is an Arlington And Other Shit district, as I referred to it the first time I wrote about it - which has sizable Black (20%) and Hispanic (22%) populations. This district was Cruz +3 and Trump +3, but while the Tarrant portion of the district barely moved, from Beto +11.5% to Biden +11.9%, that elides a lot of the shift under the hood, with Beto doing better in the urban Arlington areas while Biden did better in the white suburbs, a fact that should surprise nobody. None of this is a shock.
The district contains a bit of the DFW quad - the bottom right corner of Tarrant, and this map from Jackson Bryman shows how the very minimal topline swing is actually two counterbalancing swings, as it is in the whole of the DFW Quad.
Now, I know what you'll be saying - a district that's 52% white by population will be more white than that when you apply a voter screen on it, and I don't disagree. Echelon Insights released some electorate composition projections before 2020 in a handful of Congressional Districts, and their screen moved the (similarly ethnically diverse) Texas 22nd about 10% points whiter when comparing populations to electorates, which would make the 6th about 62% white, give or take. Seems reasonable enough to me, maybe a bit high if you think that Trumpian low-propensity whites and Hispanic don't turn out, maybe a bit low if Black turnout sags. But yeah, something like a 60-65% white electorate would be reasonable.
This poll was 75% white.
I mean, honestly, the conversation about trusting polls in general should be had anyways, but this poll, in specific, is a clusterfuck. The conversation about whether this poll is reliable should start and end with the fact it was commissioned by the Free Beacon, but even beyond that, the overly white sample is a tell that it's overstating Republicans. But let's ignore that for a second, and look at the recalled Presidential vote - Trump +3. Given it's a Trump +3 district, great job, right? Nope.
Recalled past vote always skews to the winner of the election, as more people remember voting for him than actually did. You saw that in the UK in 2019 a ton, as polls would show recalled past votes with the Tories wildly ahead despite the Tories only winning the 2017 election by 2% in the popular vote. It was a favoured way to unskew the polls, but it never held up to any scrutiny, because of the well known problems with recalled past vote. So, getting a Trump +3 recalled vote, far from being a sign of good polling methods, means the sample is probably too Republican-friendly … which we already know because it's too white of a sample. This tweet from Nate Cohn seems on point here, given the mediocre nature of the sample.
So, what's the actual state of play in the Texas 6th? Democrats will presumably make the runoff with Jana Lynne Sanchez, the GOP will get one of their potential nominees through, and Democrats are still the underdogs to actually flip the seat, but not out of the game by any means. This poll was R+10 when they asked just a generic D/R ballot test, which would represent a 2% swing to the GOP, but this is an overly white sample from a GOP pollster, so my prior - a swing to the Democrats from the 2020 Congressional result and a better result for the GOP as compared to the Presidential - is still the likeliest outcome.
There is no evidence of Democratic doom here, neither in terms of the 2021 special election, or the 2022 midterms. What we actually have is people not learning any of the lessons of 2020, trusting anything that calls itself a poll blindly, and not remembering that polls can - and often do - entirely miss the mark. Maybe this poll is right, but it will be right in the same way I was right when I wrote that Matt Bevin's career in politics was over weeks out from the 2019 Kentucky Governor's election. I was right, but fuck I was lucky, because I got to the right answer the totally wrong way.
We all seem to know cognitively that not all polls are created equally, not all polls are good, and even the "good" polls can be shit. And yet, whenever a poll comes out showing the Democrats in a potentially bad spot, the doom comes. The chances of a Democratic lockout in Texas 6th are very low, and the chances are they'll probably outperform their November 2020 results by a bit, even based on today's poll.