Sunak and Starmer reaction to MRP polling

With a General Election at most 12 months away, and potentially as soon as May, any information about the composition of the next Parliament is essential for public affairs work and campaigning.

This is why the MRP (multi-level regression and post-stratification) polling from YouGov released by The Telegraph this week has made an immediate splash.

MRP polling is a way to use large samples to predict public opinion at a more granular, constituency level. People vote based on multiple characteristics, such as age, gender, and occupation. This specific combination of characteristics is rare, but the individual characteristics are shared by many. MRP polling looks at each criteria to get an idea of how different groups will vote. It can then calculate probabilities of a specific type of person to vote in one way and using datasets is able to import this to specific constituency demographics.

This specific MRP found that if an election was held today, Labour would win a 120 seat majority and the Conservatives would slump to just 169 seats. It predicted significant Liberal Democrat gains, to 48, and that the SNP seat share would nearly halve to just 25 seats. Significant casualties for the Conservatives would be Chancellor Jeremy Hunt, Leader of the House of Commons Penny Mordaunt, and rightwing firebrand and recent returner to the backbenches, Lee Anderson.

Remarkably, the reaction from the top echelon of the major parties was very similar. In a Whatsapp to Labour MPs, the Party warned of complacency, with Labour Leader Sir Keir Starmer encouraging members, activists, and MPs to campaign like they were five points behind. Prime Minister Rishi Sunak insisted the only poll that matters is the one on election day.

However, it is too simple to look at this poll at face value.

Firstly, there are immediate concerns about who funded it, with suspicions arising when Lord Frost was quoted in The Telegraph article saying the only way to avoid the results was ‘to be as tough as it takes on immigration, reverse the debilitating increases in tax, end the renewables tax on energy costs’. It turns out the MRP polling was funded by the Conservative British Alliance, an organisation without even a website. Speaking to Politico, polling expert Lord Hayward noted his suspicion of polls funded by less than transparent organisations.

Secondly, YouGov itself called into question much of The Telegraph’s analysis of the results. It dismissed claims that Labour’s support was only up an average of 4% across the country, saying the headline vote intention showed Labour to be at 39.5% and the Conservatives 26%. YouGov also distanced itself from the claim that there could be a hung parliament if Reform UK did not stand. In fact, a YouGov poll in October 2023 found that only 31% of Reform UK voters said that they would vote Conservative if a Reform candidate did not stand.

There are also greater questions about the effectiveness of MRP polling.

President of the Liberal Democrats Mark Pack argues that MRP polling is not specific enough to deal with some extreme individual circumstances and MRP performs far better when there are greater variations between constituencies.

Polling expert Peter Kellner takes this on further with his explanation of the defects of MRP. Kellner explains that MRP polling uses an essentially proportional model which means that the more a party wins locally in one election, the more it risks in the next one. This means that when the Government loses a significant chunk of support, MRP polls predict many more seat losses than polls that use uniform national swing (UNS), the other main way to gain polling data.

Kellner argues that often the pattern of swings in Britain is closer to UNS than proportional drops. Kellner breaks swing voters into ‘grumblers’ and ‘defectors’. Grumblers are more likely to show dissent in by-election and mid-term elections, while ‘defectors’ are those who determine a General Election result as they are willing to abandon the party in Government. This phenomenon has led to by-election defeats, with former Conservative strongholds evidently having a lot of grumblers, and leads to MRP polling leading to the Conservatives losing seats which they have large majorities in. However, Kellner argues that defectors tend to often have weaker loyalties and can likely be influenced by events, and believes that UNS is a better way to measure ‘defector’ numbers while MRP can predict by-election results. Kellner thus predicts that Labour’s lead is likely to narrow in the polls as the election approaches due to voter behaviour change, as people are no longer ‘grumblers’ but have to decide what their future Government is.

If Kellner is correct, the MRP polling from this week may be less impactful than much of the media is presenting. Significantly, 17.5% of voters in another recent YouGov poll, did not know who they would vote for at the next election. Previous studies have found that undecided Conservatives still lean towards the Party and it is also likely that the Reform UK vote will decrease once an election is called. Perhaps Starmer’s caution and Sunak’s optimism are not uncalled for.

Altogether, although the MRP polling is insightful, it is important to remember its source and that polling can not always be taken at face value. Just ask Neil Kinnock after 1992 or Theresa May in 2017.

For regular updates on what is happening in UK politics and public affairs, sign up to our weekly Point of Order newsletter, going out every Friday morning.