Dem
51
GOP
49
When we wrote about that shoddy poll on Friday, the one about how 90+% think that Joe Biden, Barack Obama and George W. Bush profited off the presidency, we presumed it would be at least a couple of weeks until we stumbled across another poll that bad. Not so much, as it turns out.
By now, many readers will have heard about the latest from The Washington Post/ABC, which gives Donald Trump a staggering 10-point lead over Joe Biden among registered voters, 52% to 42%. This is, to be blunt, crazy. Over the course of the past 5 years, there have been over 1,000 Trump-Biden polls, and Trump has enjoyed this kind of lead in.... one of those prior to the new poll. Further, there is ample evidence that Trump's ceiling is roughly 46%, since that's where he checked in for both of his elections.
The cross-tabs are even more nutty, if that is possible. The pollster, Langer Research Associates, would have you believe that if the election were held today, Biden would only win the non-white vote by 9 points—a swing of 36 points from 2020. The results also say that Trump would win voters 18-25 by 16 points. That would be a swing of 31 from 2020, when Biden won that demo by 15. There are also large swings toward Trump among rural voters (+26), independents (+26), moderates (+22), and urban dwellers (+14).
Because the costs of the poll were split by two news-gathering organizations, each of them wrote up the results. Reading them both presents an interesting dichotomy. The Langer in Langer Research Associates is Gary Langer, who was ABC's polling guy before he formed his own firm. And so, ABC let him write up the story. His analysis, such as it is, treats the numbers as very serious results with very serious implications for Joe Biden. Not surprising that he would begin with the presumption that of course his firm's findings are legit.
The Washington Post's writeup strikes a rather different tone. Indeed, it's almost apologetic. The three staffers who worked together on the piece note:
The Post-ABC poll shows Biden trailing Trump by 10 percentage points at this early stage in the election cycle, although the sizable margin of Trump's lead in this survey is significantly at odds with other public polls that show the general election contest a virtual dead heat. The difference between this poll and others, as well as the unusual makeup of Trump's and Biden's coalitions in this survey, suggest it is probably an outlier.
You think?
The thing is, the previous WaPo/ABC poll was also an outlier. Not as bad as this one, but it did have Trump up 6. It would be rather unusual to get two outliers in a row, especially with one so extreme. That suggests that there might just be something wrong with Langer's secret sauce—the firm's model of the electorate. And yes, it's true that maybe Langer is right and everyone else is wrong—that's happened before (albeit not with presidential campaigns). But again, Langer's numbers are not only at odds with every other pollster (outside, maybe, Rasmussen), but also with actual, real-world experience. There's absolutely nothing in the past two sets of election results that supports the notion that Trump could get 52% of the vote.
The poll has such a stench emanating from it that politics-watchers are having a field day tearing it apart. For example, Larry Sabato, of Sabato's Crystal Ball, tweeted:
Ignore the Washington Post - ABC poll. It's a ridiculous outlier (Trump up 10 over Biden—laughable). My question: How could you even publish a poll so absurd on its face? Will be a lingering embarrassment for you.
Nate Cohn, of The New York Times, also had some thoughts:
It's really really hard to release outlying poll results, so you've got to give credit to ABC/Post here, but I do have a fairly major quibble with ABC/Post here: if you release consecutive 'outlying' poll results -- R+7 in May, R+10 today -- you don't get to dismiss your results
If it happens twice in a row in the same race, it's clear that this is the result of some element of your approach, and either you either [sic] need to decide you're good with it and defend it or you need to go home
Cohn erred slightly; the earlier poll had it R+7 among all respondents and R+6 among registered voters, while for the newer poll it's R+9 and R+10. So, the tweet should have paired R+7 and R+9, or R+6 and R+10, since those are the equivalent pairs.
So, it's a silly poll, and we have to assume that Gary Langer is going to be having some long discussions with his two clients. Meanwhile, for at least the next few WaPo/ABC polls, everyone—including us—is going to be looking very closely at the crosstabs. (Z)