Techne UK

[Pete Whitehead]

Cotswold North now as well wtf? If anything shows what shite EC is..

[edgbaston]

Jun 28, 2024 9:11:21 GMT Pete Whitehead said:

Cotswold North now as well wtf? If anything shows what shite EC is..

I’m assuming it’s based on a four way split of votes, something like 25-25-25-15-5-5 between Ref Con LD Lab Grn and other

[edgbaston]

It does raise the question if there are any seats reform could win on a low vote share due to an even split between two other parties. Herne Bay?

[hullenedge]

[hullenedge]

[hullenedge]

[batman]

Labour will definitely take comfort from this. There has yet to be a poll showing that people are ready for the Tories to return to government - although it would be surprising if there isn't one at some point in this parliament. I am a little sceptical about some of the numbers in the poll, but they don't necessarily contradict yesterday's generally poor by-election results, which could simply be the Tories getting their voters to the polls more efficiently in the context of council elections where the government of the land is not at issue. But it's a pretty decent poll for Labour, it's not exactly honeymoon territory but it's not bad at all.

[gibbon]

Batman makes some very valid points. Despite the bile from the Express (Daily Farage) and other papers it will probably be some time before the electorate will consider returning the Conservatives to government. They face Reform and the Lib Dems on the other, both have made inroads into their natural supporters. They have to find a way to defeat both if they are to return to their former glories. Until the leadership election is over and the new Leader of the Conservatives has found a way to keep his internal opponents happy the Conservatives will struggle. Next year's local elections may show some improvement in the local elections but the real test will be in 2026.

[batman]

interesting you use the male pronoun, it could well be Badenoch even if without Strathspey

[Adam Gray]

Techne were one of the poorer pollsters (14.3% absolute error, Will Jennings) in an election all pollsters got badly wrong - yet again - but which the loveless landslide drowned out.

Has Techne explained why they got it so wrong and what they've changed in their methodology that should make us pay the slightest attention to this one? I can't find anything on their website apologising for their old dodgy polls and it very much looks like business as usual: falsely maintaining a narrative that Reform is remotely close to the Conservatives and oversampling younger, more highly educated, easier to reach Labour supporters.

It seems unlikely that anyone who stuck with the Tories in July has suddenly switched to Reform, or that Farage's rabble have surged from 12% to 18% despite being largely out of the news. How many millions of voters does that represent?! Do me a favour. If Techne reckons Labour's on 33% [and they haven't refined their methodology] they're more likely to be significantly sub-30%.

[robert1]

On the Techne website there is the opportunity to send an email to ask questions.

I asked essentially what Adam Gray mentions i.e. have they changed their processes which would make them any more accurate. If/when I receive a reply I will post the details.

[The Bishop]

A few things to note in reply to AG's post.

Firstly, all VI polls taken since the GE have shown an increase in Reform support - for whatever it is worth (which may not be a massive amount, granted)

Secondly, all pollsters will surely be weighting according to how people *actually* voted in July - which is easier to do after a few months than 4-5 years and will likely make at least their raw data more accurate. The question is then what you do with it - and here there is always the danger of "fighting the last war"; all elections are different and some aspects of the last one may not prevail in the next. This is basically why most polling firms got 2017 wrong.

[redtony]

if labours ultra targeting reduced there vote to 34% posters should compare their current poll to their last poll not the GE result

[Adam Gray]

Sept 21, 2024 8:51:08 GMT The Bishop said:

A few things to note in reply to AG's post.

Firstly, all VI polls taken since the GE have shown an increase in Reform support - for whatever it is worth (which may not be a massive amount, granted)

Secondly, all pollsters will surely be weighting according to how people *actually* voted in July - which is easier to do after a few months than 4-5 years and will likely make at least their raw data more accurate. The question is then what you do with it - and here there is always the danger of "fighting the last war"; all elections are different and some aspects of the last one may not prevail in the next. This is basically why most polling firms got 2017 wrong.

Sure, but there haven't been many polls because - delightfully - the ones that were most prevalent and most wrong [Hi there YouGov! Hi there Redfield & Wilton!] have had the grace to go away and [hopefully] work on why they were so wrong. I'd also make the point that if the methodology remains wrong, the polls showing a Reform increase - and essentially straight back to where they were wrongly claiming them to be before the GE -aren't going to instill that much confidence. 

I've only seen voting intention polls from Techne and More in Common since July; I know there have been other polls conducted but those commissioning them have either not asked VI questions or chosen not to flag them, focussing instead on questions about policy and leader ratings which are perhaps less prone to go wrong (I suspect even if Labour voters are oversampled still, plenty of Labour supporters would be more willing to be critical of Starmer for example).

I don't know if you've seen the BMG review of GE polling published a couple of weeks ago but if there is a fundamental problem (across the board) in reaching more Conservative-leaning voters because they're less likely to have mobiles or be on the internet that can't be resolved (much) by weighting because, as The Bishop says, weightings are prone to change, sometimes rapidly. There may well soon be a shy Labour phenomenon as we approach mid-term and less of a shy Tory issue. 

A poll is essentially a model and all models are prone to the subjective choices made by the person who designs the model.  That's why, personally, I welcome a dearth of VI polls so that those who wish to continue trying to poll politics can have a serious think about how to try to be more accurate. 

PS I really miss ICM polls. Such a shame they got it so wrong in 2015 (?) and quit the field.

PPS Did most pollsters get 2017 wrong?  Are you saying wrong in terms of the massive Tory lead that prompted May to call the GE [cos if so, the 2017 local elections and the Copeland by-election suggested that lead was real until she unleashed her catastrophic self on the campaign] or they ended up wrong, cos they all narrowed pretty narrowly, and the roll-out of MRPs for the first time were pretty accurate?  2017 was nowhere near as wrong as 2024.

[Sibboleth]

It was actually the 2017 election that did for ICM: they projected a 12pt Conservative lead and it was more like 3pts. Quite a horrible error given that it meant the difference between a landslide and a hung parliament. Most firms did quite badly, with only Survation and Kantar doing well. I tend to think it's a mistake to have a preference for any polling firm based on performance (i.e. in a quite different sense I like MiC more than their competitors as they aren't a commercial operation and are trying to provide data for socially useful purposes, but that doesn't make their polling automatically any better even if it means that some pressures are averted by definition), as good runs never last as, these days, they're mostly based on getting guesses right: as the fate of ICM demonstrates itself.

In any event, polling presently is only of any use as a point on a graph. A General Election is a hypothetical thing in the relatively distant future at present and there's not much enthusiasm for politics for now (just look at turnouts in local by-elections). No one can exactly be 'right' or 'wrong' as none of it is exactly real. I would agree with you that having fewer voting intention polls outside election periods is greatly preferable to the gluts seen in recent parliaments.

[The Bishop]

Sept 22, 2024 0:26:18 GMT Sibboleth said:

It was actually the 2017 election that did for ICM: they projected a 12pt Conservative lead and it was more like 3pts. Quite a horrible error given that it meant the difference between a landslide and a hung parliament. Most firms did quite badly, with only Survation and Kantar doing well

Kantar weren't around back then, can't recall their predecessor being that distinguished.

Apart from Survation (who were actually unwisely mocked for their 2017 campaign efforts by certain other pollsters, including ICM) I think YouGov were the closest - their MRP was widely disparaged but was remarkably accurate, though their regular polling efforts were marred by chickening out and "adjusting" their final pre-election effort in a way that helped the Tories. ICM made a short-lived return to the fray for the 2019 campaign but again didn't overly impress, and since seem to have vanished entirely - at least as far as political polling is concerned.

And the interesting thing is that as AG says *all* pollsters gave the Tories a massive lead when the 2017 GE was called - the issue was how they diverged subsequently, and those that went wrong were the ones who assumed things would play out as they did two years earlier.

Another point to consider is that since 2005, the only pretty good election for the pollsters has been 2019. Which shows what an unusual episode that was, if nothing else.

[batman]

Kantar were definitely around in 2017 but they probably used one of their other brand names such as TNS. I know because I was working for Kantar in 2017. They also did the System 3 poll in the Glasgow Herald.

[The Bishop]

Yes they were TNS back then. Their final effort for 2017 was 43C 38Lab which wasn't a million miles out tbf.

[hullenedge]

[robert1]

Sept 21, 2024 6:57:07 GMT robert1 said:

On the Techne website there is the opportunity to send an email to ask questions.

I asked essentially what Adam Gray mentions i.e. have they changed their processes which would make them any more accurate. If/when I receive a reply I will post the details.

I have reminded Techne that I wrote a week ago. I have yet to receive a reply.