A number of us here at DailyKos have been commenting for a while on the fact that RW pollsters have been releasing a firehose of questionable, very right-leaning polls. It’s been baffling that the lesser Nate (the one who runs 538) has simply accepted these polls without question as they pushed the predictions more and more rightward. What does he care? His current prediction for the senate is nearly 50-50, so he gets to claim he was right either way.
Well, the somewhat better Nate at the NY Times finally took note of this, and in the mildest possible tone, decided to bring it to readers’ attention (Nate S.’s tweet above is subtweeting this article). For those without a subscription, here’s the key passage:
Many stalwarts of political polling over the last decade — Monmouth University, Quinnipiac University, ABC/Washington Post, CNN/SSRS, Fox News, New York Times/Siena College, Marist College — have conducted far fewer surveys, especially in the battleground states, than they have in recent years. In some cases, these pollsters have conducted no recent polls at all.
And on the flip side, there has been a wave of polls by firms like the Trafalgar Group, Rasmussen Reports, Insider Advantage and others that have tended to produce much more Republican-friendly results than the traditional pollsters. None adhere to industry standards for transparency or data collection. In some states, nearly all of the recent polls were conducted by Republican-leaning firms.
There are a number of reasons that the RWers would do this: fundraising grift, to push the media narrative (which has certainly worked), and as a number of you have pointed out, to provide justification for violence and accusations if/when they lose the Senate, the House, or both.
And the RW polls aren’t merely right-leaning (though they are — I estimate that Trafalgar, for instance is consistently about 5-6 points to the right of, say, Siena or Marist — but that they are regularly employed by Republican firms, put a RW spin on press announcements, select their polling targets and timing to maximize their damage to Dems, etc. There is simply no good reason to assume that they are operating in good faith (and only minimal reason to suppose that they’ve actually conducted a poll at all rather than simply made up a number). More to the point, the only ones with significant track records (longer than 2020) are Rasmussen and Trafalgar who missed the mark in the rightward directions in both 2018 and 2020 (and in the intervening specials and off-year elections) by 2-5 points on average. The rest of them have clustered with them.
Simon Rosenberg has been going on about this on twitter, and suggests it’s incumbent on us (technically the press, but I’m not going to hold my breath) to produce more realistic averages:
Not unskewing, just average only truly independent, media affiliated pollsters. So, okay, let’s do that. Using only: Siena, Hart, Quinnipiac, Monmouth, Marist, YouGov (CBS), Beacon Shaw, Marquette, Selzer, Suffolk and Ipsos from the last 3 weeks, we get the following for competitive races:
PA: D+4.3 (5 polls)
AZ: D+4.0 (3 polls)
GA: D+3.0 (4 polls)
NV: even (3 polls)
NC: even (1 poll)
OH: R+0.5 (2 polls)
WI: R+2.0 (3 polls)
You can take issues with my pollster choices if you like (and I’m happy to field specific questions for additions/deletions), but in the absence of polling errors, we’d get 50 seats free and clear, and up to 53 if the next 3 break our way. Some of these are almost criminally under-polled, and the press is treating them as a given that they’ll go R. Oh, and for what it’s worth, all 5 of the PA polls were taken at least partially post debate.
Finally, FWIW, in 2018, the average error of all of these pollsters who did multiple final weeks polls was about 0.2 points (toward the Dems, if you must know). Sure, they did lousy in 2020, so if your working assumption is that we should just add 5 points to the Republicans in all polling averages, I can’t stop you, but it requires thinking that there weren’t special challenges 2 years ago (DeJoy, Trump on the ballot, pandemic voting, etc.) and that none of the pollsters have adjusted their approaches at all in the interim. Supposing this year will be just like 2020, in light of Dobbs and the many elections in the last couple of years seems unnecessarily pessimistic if you ask me.