In 1992, every opinion poll and market research organization predicted a Labour victory at the general election. The BBC exit poll as the polls closed at 10pm on 9 April 1992 suggested that the Conservatives “would only just get more seats than Labour”. They were all wrong. John Major and the Conservatives won with a modest but clear overall majority of 20. The Conservatives vote share went down by a mere 0.3% from the previous 1987 election. The Conservatives vote share was 41.9%, compared to Labour’s 34.4%. Conservative MPs totalled 326: Labour 271. A difference of 7.5%, of 55 seats, is not “only just ..more…than Labour”.
So how did the polls get it so wrong? The initial response from the polling organisations was that there must have been a last minute swing. So last minute that the exit poll didn’t even pick it up? That didn’t wash. And then, and then, two organisations suggested that people might have been lying in their responses. As own goals go, this was spectacular. The market research and polling organizations were telling the world, including their own private sector clients – that the results of their expensive market research were unreliable.
“Eight out of ten cats prefer Whiskas.”
Consumer market research is longstanding. Traditionally, a common method was for an independent market research organization to ‘blind test’ products, stripped of branding and presentation, to gauge preferences. Most famously, this led to the manufacturer of a certain petfood being able to claim, with some evidence, that ‘Eight out of ten cats prefer Whiskas’. But think of this – what if one out of ten cats had preferred Whiskas? Then the market research report would have been quietly binned: perhaps we would never have heard the famous strapline. So how many opinion polls might be unpublished because the commissioners of the report don’t want the results to get out? Certainly major political parties have ‘in house’ polling which is rarely reported – even if it were, it would lack some credibility because of the source. But others have remained unpublished as well.
Polling organizations are often profit making market research companies – this was their origin and their motives remain the same. So clients/customers pay for polling to be carried out. Some other polling organizations have different goals – let’s examine some of them.
Firstly ‘More in Common’, set up as a non profit organization after the killing of Jo Cox in 2016.
From the depths of their website, we find this. “More in Common UK is funded by a mix of our consultancy work and institutional funders. In the calendar year 2025, More in Common received income from 80 separate organisations, with no single organisation accounting for more than 10% of our income. Around two thirds of our funding comes from consultancy projects with businesses, campaign groups, charities, local government, and non-profits. The remaining third of our income comes from institutional funders.”
The site also states “Our hope is that through our work, we honour Jo’s memory and legacy”. Now, given that Jo Cox was a Labour MP, that has presumably to be a Labour oriented legacy. It may well be the case that More in Common tries to be, or even is, impartial. But it is difficult to avoid the perception that this is an organization with a tendency towards Labour. Would this influence people’s responses?
And why would institutional funders wish to subsidise an organization to the extent of one third of it income when other very similar organizations exist?
Secondly, ‘Yougov’. From the website, we find “YouGov offers the most accurate data, engaged panel members, pioneering technologies, and trusted experts.” The panel members are part of an online community – Yougov is entirely internet based. A private company, Yougov was founded by a small group of shareholders in 2017, including former Conservative party Chairman and now Reform MP Nadim Zahawi. Again, the organization may well be, or certainly try to be, impartial, but the clear connection to a right leaning politician may influence perceptions.
Thirdly, Lord Ashcroft polls – Lord Ashcroft is a former Deputy Chairman of the Conservative party. His polling is largely self funded and, in fairness, his name is on the label so there is no doubt as to the source.
Fourthly, Convergent. Founded by two people who worked ‘for the Labour Party through the 2024 election’, who say “We both started by working at leading traditional polling agencies. We then built the Labour Party’s in-house polling and data science capabilities — and helped deliver the UK’s biggest election victory of the 21st century, with more accurate predictions than any public pollster. We’ve since run polling for the top of the UK Government and No.10.”
Get the last sentence – having worked for the Labour party pre-election, they are now paid by the government to run polling.
Here are the latest opinion polls from each organization.
| Date(s) conducted |
Pollster | Client | Area | Sample size |
Lab | Con | Ref | RB | Lead |
| 2–3 Aug | YouGov[7] | The Times/Sky News | GB | 2,300 | 22% | 19% | 23% | 4% | 1 |
| 30 Jul – 3 Aug | Lord Ashcroft Polls[9][b][a] | N/A | GB | 5,303 | 27% | 23% | 18% | — | 4 |
| 31 Jul – 2 Aug | More in Common[10] | N/A | GB | 3,334 | 27% | 22% | 24% | — | 3 |
| 25–30 Jul | Convergent Opinion (MRP)[13] | N/A | GB | 10,919 | 26% | 21% | 23% | — | 3 |
Source: Wikipedia 12/8/26
A few comments. Yougov is the only one of the four to give Reform a lead. The Lord Ashcroft poll is the most favourable to the Conservatives in terms of vote share. More in Common and Convergent are very similar and show a three point Labour lead.
One would not suggest that these polls vary because of the political lean of the organization – but there is evidence that respondents ‘tailor’ there response according to who is asking the question. Convergent (and others) claim to deal with this by adjusting the results they get to account for perceived under/over representation of differing demographics in their responses. This supposedly allows the results to be applied to individual constituencies. The method is MRP (Multiple level regression and post stratification) and is well explained here: https://britpolls.co.uk/explainers/what-is-mrp/
Note also that Convergent and More in Common state that there is no client. But do remember where their funding comes from.
And also note that only one of the four polls included Restore in their responses. I deliberately omitted minor parties such as Green/SNP/Liberals from the diagram for reasons of presentation but include Restore only to make the point about which parties pollsters include or do not include.
So what are opinion polls for? To reflect opinion? To forecast behaviour by adjusting results through methods such as MRP? To provide a good headline for a political party? Or even to shape opinion?
To be continued.
© Hongkonger 2026
