This was a Democratic party primary, but there have also been a lot of big misses in general election polling recently, consistently underestimating Democratic support.
I continue to suspect that there is some segment of Democratic voters that polling is largely failing to capture - my hypothesis has been young, very progressive, and very internet savvy voters, that I dub “Swifties”, are particularly adept at avoiding pollsters, whether by phone or online, and are consistently under weighted in polling. I wonder if we could look at the details of the above primary - perhaps Alsobrooks (the winner) had the lion’s share of the Swifties, and that’s why she was consistently underestimated?
If I’m correct, then if Biden gets the Swifties, he’s likely to over perform the polls.
This isn’t a polling issue per se, but given the severity of the consequences of Trump winning, polling has to kind of take that into account. A doctor telling you that there’s a 50% chance of you having the flu is drastically different than a doctor telling you that you have a 50% chance of having pancreatic cancer, even if the percentage is the same.
In other words, a 4% Biden lead in the polls over Trump has to be considered a lot less reassuring than, say, a 4% Biden lead over Romney or a normal Republican.
It’s not about motive for pollsters - they’re just having a harder time getting the Swifties (if my hypothesis is correct). The Swifties use cell phones, and are more adept at avoiding callers they don’t want to talk to, and avoiding online solicitations they’re not interested in. They have other things they want to do with their phones.
[The New York Times] released a second national poll (! -b) conducted with a different pollster, Ipsos. This poll did not find the kind of structural weakness for Biden their big Monday poll did. In this poll Biden leads 45.5% to 42.5% (+3) in their initial head to heads between the two candidates.
To understand what an outlier Monday’s NYT battleground states poll was let’s look at the other national and state polls released in recent days (Biden-Trump):
NPR/Marist 51%-46% (+5) likely voters
ABC News 49%-45% (+4) likely voters
NYT/Ipsos 45.5%-42.5% (+3)
RMG 44%-42% (+2) likely voters
TIPP 42%-40% (+2)
Big Village 42%-41% (+1) likely voters
Data For Progress 47%-46% (+1)
Economist/YouGov 43%-43% (tied)
Yahoo/YouGov 45%-45% (tied)
USA/Suffolk 37%-37% (tied)
Morning Consult 43%-44% (-1) and 43%-43% (tied)
Quinnipiac Wisconsin 50%-44% (+6)
Orbital Digital Arizona 39%-38% (+1)
CBS News MI 51%-49% PA 49%-50% WI 49%-50%
12 national polls. Biden leads in 7, 4 are tied, Trump leads in 1. State polls by serious credible pollsters showing Biden in far better shape than the NYT polls. All of these polls, with the exception of Data For Progress are independent polls, not controlled or paid for left of center organizations. I include DfP because historically their polling has not been favorable to Democrats.
Part of the reason I got the 2022 election right when so many got it wrong was that I broke out independent polls like these from polling averages which had been influenced by a flood of Republican aligned polls. These Republican funded polls often produced results 3-4 more points more Republican than independent polling. I think we are starting to see something similar happen in this election, with most independent polls finding a close competitive election, perhaps one now where we have a slight advantage; and R funded polling (and NYT/Siena) finding it much more Republican. It’s clear Trump and Republicans think they are seeing a new red wave, and that they lead. I don’t think it’s a fair read of the data, just as I didn’t think the red wave was a fair read of the data in 2022.
What are “republican aligned polls” and polls “controlled by left of center organizations”?
How do their methods differ from " normal" polls?
I can see why they mighr choose to publicize or not publicize certain results of their polls. But how does their methodology differ, so that it consistently creates the results which they prefer?
I think this needs to be generalized. Yes, the pollsters are missing your “swifties”, but who’s to say that there aren’t a large number of young, conservative, internet savvy voters (in my mind, modern yuppies) who are adept at avoiding pollsters.
Lord knows that the pollsters aren’t intentionally missing the “swifties”; the swifties might not be avoiding the polls intentionally, it’s just a side effect of their lifestyle - the polls just need to find them.
Society has changed; the polls are changing, but probably not fast enough. As we (in general) change our habits, I have faith in the long run that the statisticians will find the correct mix; probably just in time for society to change again.
Biased polls. If you see a poll where the first question is “Do you approve of Biden’s open border policy for illegal aliens?” it’s probably a “republican aligned” poll. The questions are skewed to elicit a certain response. Many of them are much more blatant that that.
This is possible too, though ISTM that younger and tech savvier people are overwhelmingly progressive, but I don’t have any data at my fingertips to support this.
Lots of segments of people are unlikely to participate in polling. Pollsters aren’t stupid though and I’m not sure why so many people here think otherwise. At least the good ones understand that there are groups that they will get a limited sample of and take steps to give them proper sampling.
Polling isn’t perfect, and individual polls have limited value, but the aggregate polling figures are the best evidence we have of the state of the elections. Certainly way more accurate than TV pundits or esoteric message boards. Even in the age of ubiquitous cell phones, there is no evidence that polling is getting less accurate. There also isn’t one side that the data routinely over or under counts. Often a pollster will correct for errors from prior elections and the new error will be based on if they over or under corrected.
If I’m the Democratic I trust the polling is more or less accurate and stop trying to live in denial and stop blaming the media for delivering news I don’t like. This is absolutely a winnable election, but the polling is showing real problems for Democrats with minorities and the youth and courting those groups needs to be a primary focus.
You can see that it is correct to say that the average had Trone ahead, but there weren’t really enough polls to state a good average. And the latest poll had Alsobrooks ahead by 3. I would argue that the earliest polls are pretty useless due to the high number of undecided voters.
The only one that was way off was the Baltimore Sun poll from early April that had Trone up 12. That seems to be an outlier. It was done by an outfit called OpinionWorks that I’ve personally never heard of. It has a terrible rating from 538 (1.8 stars).
So the rule of thumb remains: take an average of lots of reputable polls, and accept that there may be a swing in either direction.
Again you have to look at things in the aggregate. The fact that their are some elections that were off one way, does not mean that either polling has become more inaccurate or that their is a bias in a certain direction. You don’t have to go back very far to see elections undercounting republicans.
Another example of sampling bias was a March 2024 Grinnell College/Selzer Trump +7 poll that was rolled into the RealClear Polling average during late March and early April. Scroll down to the section titled “Biden Trails Trump and Loses Ground in Key Demographics” and the caption of the graph therein reads:
Source: Grinnell College National Poll - conducted by Selzer & Co.,
of 1,005 U.S. adults including 715 likely voters (308 Republicans, 231 Democrats, 158 Independents), March 2024.
…
It’s also possible to run a squeaky-clean poll (or nearly so) in earnest, and then to repeat the process a few more times. A pollster can then pick just one set of results to report publicly – one that best supports the narrative of their choice.
Worth noting in the MD case that Trone is a billionaire who owns a chain of 250 big-box liquor stores, whereas Alsobrooks is an African-American civil rights lawyer and longtime ally of Obama and Kamala Harris. In other words, Trone is exactly the kind of business-friendly candidate that the polls seem to be skewed in favor of these days.
One area of how polling might not reach the younger sector are indeed smartphones. I can’t speak for Android, but I do know that since fairly recently, iPhones screen calls based on your contact list. “Silence Unknown Callers” is on by default. According to one survey, nearly 90% of Gen Z Americans use iPhones. So, if cold calling is one of the major tactics in reaching potential voters for most pollsters… they ain’t getting through.