This is very good from Peter - and basically shows how the British Polling Council rules are supposed to work. kellnerp.substack.com/p/how-whiteh...

Why a media scoop about Labour's internal polling is less exciting than that makes it sound - and also some rather strange things in the details of how a poll about striking doctors has been covered buff.ly/oK0gbtu

I am not impressed by the polling some people keep sending me on assisted dying: buff.ly/wErlwoC

There's going to be a lot of focus on the topline -- that polling was more accurate than in 2016 and 2020 but still biased towards Dems on average by 2.7 pp -- but I want to highlight some of the other findings I found most interesting

We either exclude or reassign undecideds in voting intention polls because the purpose is to get a picture of how people would vote, and 'don't know' isn't an option in the voting booth. I don't see a justification for it when being uncertain, conflicted, or uninformed is a valid position.

Great blog dissecting a badly conducted, badly presented and badly reported poll. Stuff like this damages the credibility of polling (and of poll reporting).

Shocked, SHOCKED that this person is doing this thing again

My latest Substack examines how flawed polling and flawed journalism are distorting the debate over assisted dying kellnerp.substack.com/p/peers-beware

A very good read.

I am very much of the "you should publish null findings" school of thought. This isn't really that, but as you might imagine the original inspiration of this was the media fuss around Adolescence, and horror stories about what young men thought. And, well, see for yourself bsky.app/profile/youg...

This is my regular reminder to everyone that jstor is open to the general public now; a free account there will give you access to 100 papers a year.

Absolute must-read from @profjanegreen.bsky.social and @martamiori.bsky.social. It's hard to imagine how Labour could be getting it more wrong - not least by talking up the replacement of the Conservatives by Reform, which will only accelerate the cannibalisation of the former's vote by the latter!

The actual findings of this poll aren't as overwhelming as the headline and opening paragraphs make out. The 67% who supposedly support criminalising buying sex is reached by excluding the 36% of respondents who neither support nor oppose the policy. Only 43% actually support it.

Some lovely polling published by the Policy Inst at KCL today on partial quotations www.kcl.ac.uk/policy-insti...

Feel like we should bring back the old twitter follow friday to boost good people on here who maybe didn't get an initial boost from the starter packs. So follow @psurridge.bsky.social and @profjanegreen.bsky.social - two excellent and thoughtful political scientists.

These estimates of public support for violence (from www.nytimes.com/2025/09/16/o...) are inflated - *much* higher than what we and others have found using question wording that reduces acquiescence bias: brightlinewatch.org/accelerated-... The vast majority of Americans reject political violence.

would have been good to see some push back from edsall on this. it's incredibly important to get these numbers right, and this unfortunately does not pass muster

Peter Mandelson Sacking Day seems to come around sooner every decade. My SKETCH of Today in "No! Not Peter! Not again?"

Ever struggle to explain why your research is relevant? Sometimes the data just does it for you! 👇 Has anyone seen a more dramatic poll discrepancy than this one from Spain nationwide polling? CIS has PSOE +9.0pp, while NCReport has PP +9.4pp *over the exact same fieldwork*! 🤯😅

An interesting analysis – which isn’t reflected in ‘would you consider voting for X’ data (which suggests Reform *is* reaching parts UKIP didn’t). There’s no contradiction: Reform is reaching more voters, but only because of Con/Lab collapse. Suggests ‘Reform Modernisation’ narrative is overdone.

Somewhat sensationalist headline here obscures some really interesting research - which tallies with other available tracking data - on shifting attitudes to climate in the UK. Below are a few quick thoughts on how to understand this, in my view 👇 t.co/gDqlQgQBHn

🚨NEW DATA 🚨 The BES team are pleased to announce the release of the 2024 Random Probability Survey Release v1.0.0 You can download the data here: www.britishelectionstudy.com/data-object/... You can find our release note here: www.britishelectionstudy.com/uncategorize...

Sigh. "Might consider doing x" does not equal "set to do x".

In polling world, “synthetic sampling” (using LLMs to replace human respondents) has been getting a lot of attention recently. Last wk, Ipsos announced a synthetic panel! But does it actually work? Here’s a white paper I wrote: “Your Polls on ChatGPT” report.verasight.io/synthetic-sa...

Had it been presented with comment, that comment might have noted the good reasons not to take these answers at face value. In hotly politicised contexts, poll responses are often expressive rather than sincere: that is, people give what they know is the right answer to support their side. (1/3)

"And the award for the least truthful or effective use of polling in an election campaign in all time goes to....."

The headline here should say: "Number of private schools increases by 35 - showing fears of an exodus to be unfounded". (The increase is mentioned way down the article.) www.thetimes.com/article/19ef...

The shared conviction of everyone running election night coverage that no one staying up all night to watch the election could really actually be interested in the, y'know, election is one of the mysteries of senior TV management.

Welsh election law now requires local candidates to list their party memberships in the last year. That's good for transparency re. those listed as independents. It also gives us things like this... (Abermaw, Gwynedd by-election, polling this week)

I won't be posting any opinions how ever much you like this post.

Since people keep asking... We don't know with confidence how 16 and 17 year olds would vote. Polling evidence is quite scant because young people are some of the hardest to reach (and I would expect those you can reach will be disproportionately engaged and, therefore, not very representative). 🧵

Anthony's listed a litany of VI polls with hypothetical parties that massively inflated that party's support. I'd add to it that Panelbase found that 12% of voters would "definitely" vote for a hypothetical, Alex Salmond-led party. Alba won 1.66%.

There are some hypothetical VI polls with a new Corbyn party doing the rounds. Take them with a pinch of salt. It's a very natural thing to ask, but it's very difficult to do (and these things - voting intention polls asking about hypothetical new parties - have a *terrible* track record).

lol the clear implication of this briefing is that someone associated with a PR firm pretended to be an irate rich person on the phone to get a plug for a client who sells services aimed at irate rich people

I recognize this is a losing battle, but to keep blaring that trumpet: "Push poll" is a term of art, referring to a voter-persuasion technique masquerading as a poll. That is DIFFERENT from a real poll that asks people how their vote would change if they knew [X].

I'm reading the report behind these headlines about glyphosate and it may be the most scientifically illiterate thing I've come across (although it's a close thing with the recent MAHA report). 1/n

"if a piece of writing reports poll results out to the tenths place, stop reading it" is certainly not infallible, but I wouldn't call it a terrible rough heuristic for spotting writing that fundamentally misunderstands or doesn't care about how polling data works or what makes it useful

Note the BBC Projected National Share (PNS) also includes adjustments for multimember wards to count voters equally. PNS projects to places in Britain without elections last week, and estimates what the result would have been if all 5 parties (Con, Lab, LD, Green and Reform) had stood everywhere.

% vote from 2025 local elections, with adjustments for multimember wards so voters count equally. Reform 31 Con 23 LD 17 Lab 14 Green 9 All 1399 divisions Source: BBC For info on unrepresentative character of places with elections: electionsetc.com/2025/05/01/u...

Grrr. Lowest ever level *from More in Common* (it is good practice to compare like to like, but it does mean headlines like this are rather easy to come by for newer companies). Labour last down at this sort of level in 2019 before More in Common were polling.

Very good corrective to the frankly psychopathic right-wing war on Motability exploding across twitter and the hard-right media over the last couple of weeks www.theguardian.com/world/2025/m...

What a bizarre and terrible column this is by Simon Kelner, suggesting that Starmer raising the issue of the Ebbsfleet Spiders is like Trump's claims about transgender mice. When it isn't, because the spiders thing is *actually true*. inews.co.uk/opinion/keir...

Sometimes a co-author might be listed first due to alphabetical order, or the toss of a coin. And sometimes — as in this case — the first author was the undoubted star of the show and did far more than his fair share of the work. So I’m adding my grateful congrats here!

Anyway, if MailOnline are going to stick their story behind a paywall then I'm going to make mine available for free for 24 hours. So have a read, stick this in your WhatsApp group, and just help me spread the word about where to get the real thing. www.londoncentric.media/p/harry-pott...

One for those who like field experiments with validated behavioural outcomes and political identities doi.org/10.1086/730718

Politics Live just referring to this poll claiming alarming decline in support for democracy among the young. Except there isn’t. This has now been debunked several times (Mark Pack’s linked post is the most thorough) but the only thing remembered and cited in debate is the original headline.

I have a new article in @electoralstudies.bsky.social 🎉 It uses conjoint experiment in the UK to test how important the culture war (statues, LGBT+ representation in pop culture, DEI, transwomen athletes, curriculum diversity, and university free speech) is to voters Link👇 & results 🧵

This is very good and thorough on that Channel 4 poll claiming half of Gen Z think we’d be better off with a dictator.

Very useful piece on whether "Gen Z" has really given up on democracy. It brings out a key problem with the reporting of polls: that what we often get are very contestable analyses, spiced with numbers that give an impression of scientific precision. And it takes a lot of digging to challenge that

There's more from that shonky Zoopla poll. Deary me.