This week’s YouGov/Sky News/Times voting intention poll has the following headline results:
• Reform UK 25% (+2)
• Green 19% (nc)
• Conservatives 17% (-2)
• Labour 17% (nc)
• Liberal Democrats 14% (nc)
The poll was taken on Sunday 15 March and Monday 16 March, with a sample of 2,329 respondents from YouGov’s online panel.
This is the first poll since Nigel Farage and Reform UK publicly challenged YouGov’s approach, and the pollster agreed to supply more underlying data about each poll.
How YouGov carries out voting intention polls
During the last election campaign in 2024, YouGov changed how it conducts its weekly voting intention polls in an attempt to pick up tactical voting that has become an increasing feature of UK elections in recent years.
It has continued to use this methodology since the election.
Unlike other pollsters, YouGov’s methodology involves asking its online panel two voting intention questions:
• How they would vote if a general election was held tomorrow
• How they would vote in a general election if they were thinking specifically about their own constituency
The results are then put through an MRP model (or, to give it its full name, a “multi-level regression and post-stratification” model) to turn the raw data into headline voting intention – the figures Sky News reports each week.
YouGov uses these two techniques – a pair of voting intention questions and then putting the results through an MRP model – because it believes this allows it to get the closest to the result of an election held tomorrow.
There appear to be significant differences between pollsters in their respective treatment of Reform UK: there are a lot of irregular voters currently telling pollsters they will go out and vote for Nigel Farage’s party in an election tomorrow, and different companies take a different view on how likely this would be to happen in practice.
Why Reform UK disputes the methodology
In recent months, YouGov has reported lower polling shares for Reform UK than other firms – although other pollsters also reported a decline from its peak – and Mr Farage’s party has challenged the pollster’s methodology.
It says it believes the first voting intention question – that makes no reference to constituencies – is a better representation of what is happening in the country, as well as questioning the use of the YouGov MRP model.
It points to the pollster Peter Kellner, a one-time employee of YouGov, who said the use of a second voting intention question about how a respondent would vote if thinking about their constituency would advantage the Liberal Democrats over Reform UK.
So what now?
From this week, following the Reform UK challenge, YouGov has agreed to publish the results to the question without the constituency prompt, as well as the one with the prompt, which was already automatically part of the data.
Mr Farage is claiming this as a victory for transparency. YouGov’s methodology, however, has not changed, and it stands by its approach.
So here are YouGov’s raw voting intention numbers this week without a constituency prompt, and before YouGov applies the MRP model:
• Reform UK 19%
• Green 16%
• Conservative 11%
• Labour 11%
• Lib Dems 7%
• SNP 2%
• Plaid 1%
• Other 4%
• Would not vote 10%
• Don’t know 15%
• Refused to say 3%
These are the numbers Reform UK says are the “real” figures, which each week it is likely to highlight.
Note the figure here for Reform UK is the same this week when the question is asked both with and without the constituency prompt – 19%.
Who is right?
All pollsters use modelling and a range of techniques to generate the headline voting intention they believe best reflects reality.
Ultimately, these results can only be tested at a general election, and at these moments, polling companies are judged by clients and shareholders.
This wait can be frustrating for political parties, since in between elections, polls drive momentum and, at worst, can be used to justify a change of leader.
However, at the last election, the final YouGov MRP poll put Reform UK on 15%, the exact number it received at the ballot box, and the final MRP was the most accurate by seats of any pollster, with 92% of constituencies called correctly.
