It’s been a lighter week for new voting intention polls, with only four with fieldwork since the last round-up. But they continue the picture of different pollsters, with differing methodologies, all painting a similar picture: Labour on 36-40%, Conservatives on 31-34%, Lib Dems on 10-14% and Greens on 5-8%.

Worth noting that within that picture of similarity, Opinium was the one most different from the pack (and with the lowest Labour lead, at 3%). This might be a return to the previous pattern of Opinium showing different figures from most other pollsters, or it might be just a blip. Hold your judgement for another couple of polls from them.

We’re coming up to the six month anniversary of the last time the Conservatives were ahead in a national voting intention poll and their current level of support is around that which Corbyn’s Labour got in 2019 (33%).

So should we expect a crushing defeat for the Conservatives at the next general election? Well, two MRP-based studies out this week both suggested that, with varying credibility.

First, there’s a YouGov MRP model, looking at 88 Conservative-Labour contests:

From the 88 constituencies which the Conservatives either (a) won from Labour in 2019, or (b) currently hold with a majority of less than 15pts over Labour, our modelling predicts that just three would remain in Conservative hands: Ashfield, Bassetlaw, and Dudley North.

Alas, the published YouGov data only covers these 88 seats, so for all the cleverness of MRP, we’re not able to draw any lessons about whether these seats are behaving differently from the rest of the country. It’s a dramatic echo of those national polls showing the Conservatives down at Corbyn-levels of support, but not one that adds much illumination to what the national polls tell us. Though it did give us a pollster barney over rival embargo times.

Then we have the MRP done by Electoral Calculus and Find Out Now for the Constitution Society, claiming that, “a deal between Labour, the Liberal Democrats and the Green Party at the next election could lead to an opposition landslide, with the Conservatives losing up to two-thirds of their seats”.

Their figures put the Conservatives, in the face of such a deal, at just 101 seats. To give that some context, when Labour was smashed in 1983, they got 209 seats. When the Tories were smashed in 1997, they got 165 seats. In Labour’s 1945 landslide the Tories got 197 seats. (All with a similar sized House of Commons.)

And so when I first read the press release I assumed either I’d misunderstood the table or it had a typo. But no, 101 seats it is.

Is that credible? Let’s run through some standard checks on a poll’s credibility.

First, the pollsters. Find Out Now are members of the relevant trade regulatory bodies and Electoral Calculus’s elections data has long been widely used. However, they have teamed up twice in recent years for local election predictions, and in both cases (2021 and 2022 May elections) their headline seat number predictions got the overall picture significantly wrong, e.g. for 2022 predicting Labour would gain 16 times as many seats as the Lib Dems.

Local election polling and predictions do seem to be harder than general elections, so that shouldn’t be held against them too much. But it is a reason for caution, especially as they’ve not yet had a general election prediction put to the test by a real contest then occurring.

Second, how the findings compare with the evidence from other sources. History points to their figures being very unusual. But then record books wouldn’t need updating if the very unusual never happened. We do though also have the recent Best for Britain MRP polling, also looking to capture the impact of a possible election deal. I looked at this last week, and one of its details is how few seats it actually points to switching hands because of a deal. It didn’t point to anything like a history-book-breaking 101 MPs result. I can’t think of any other data that points towards a picture similar to that of the Constitution Society. It’s polling is very much the outlier.

Third, how credible do the details look? I won’t repeat last time’s summary of the problems with polling that tries to identify the impact of seat deals but instead let’s look at the seat deals it envisages.

Take the Hampshire East constituency in which the Conservatives secured 59% in 2019, Lib Dems 24%, Labour 11% and Greens 5% (just under, losing their deposit). The poll shows the Conservatives losing the seat – though with them starting on 59%, clearly that can’t all be about a seat deal. But who do you think they lose the seat to in the Constitution Society’s model?

Lib Dems? No.

Labour? No.

Greens? Yes, really.

The Constitution Society model is that Lib Dems and Labour stand down for the deposit-losing fourth placed Greens, who then sweep to victory. It’s not obvious that this is plausible. Good luck to anyone heading off to Lib Dem or Labour HQs to pitch the ‘please stand down for the Greens here’ message.

This example is not alone in the model. Although, to be fair, there is a reason for them. It’s that the Greens are not credible challengers to the Conservatives in any Parliamentary seats. So any seat deal involving the Greens has to either involve the Greens giving up lots in return for almost nothing, or involve the Greens being given the opportunity to run solo in seats where they are extremely weak.

So the Constitution Society should be given credit for showing how such a deal might work. That’s certainly preferable than the approach of some who just assume away this basic problem with and for the Green. But the resulting model, rather than demonstrating the power of seat deals, illustrates their inherent difficulty.

(For more on judging their work, see my piece on their similar projection made last year.)

Meanwhile, in less eyebrow-raising polling in the last week:

• Although the Sue Gray report doesn’t seem to have directly significantly worsened the polling numbers for Boris Johnson and the Conservatives, the public’s view is firmly negative, as with YouGov finding that only 15% think he is “genuinely sorry about what happened in Downing Street during lockdowns”.
• A big majority of the public also continue to want the PM to resign. The figures across different pollsters have not got worse on this measure in recent weeks, and may even have got slightly better. But the problem for the Conservatives is that the figures remain dire for the PM.
• Labour also continue to poll better than the Conservatives as being the party for most/ordinary people, as with Opinium’s finding that 44% say they are the party “most on the side of people like you”, compared with only 25% for the Conservatives.
• Survation’s polling showing Red Wall voters support many (centre) left policy positions continued, with this week bringing polling showing they support large increases in the minimum wage.
• Seven-in-ten voters believe leaving the EU has made life in the UK more expensive – although that hasn’t, so far, translated into a significant, sustained shift in opinion on whether Brexit was right or should be undone.
• Yet cost of living continues to dominate voters’ concerns, with inflation reaching a 40-year in the Ipsos MORI issues index (and hooray for pollsters asking the same question, with the same wording, decade after decade) and a shocking 1-in-5 saying they either “often struggle to make ends meet” or “often have to go without essentials like food and heating.

For the full story, and the fun story, about political polling check out Polling Unpacked: the history, uses and abuses of political opinion polls. According to the Sunday Times, it is “Essential reading for anyone seeking to understand modern politics … comprehensive yet surprisingly fun”.