Giant MRP polls tell stories of Tory catastrophe – or mere disaster

This week, three polls predicting the result of the UK general election came out within a roughly 24-hour period. One predicted that the Conservatives would lose more than half their seats, ending up with 155. Another said the party would finish with just 53.

So what? That spread represents the difference between a terrible Tory defeat and the party’s worst ever result, says Sir John Curtice, the UK’s foremost polling expert. Equally, the polls might all be wrong. 

• All three so-called MRP polls sample a much larger pool of voters than normal opinion polls.

• All three map voting intentions against census-based demographic data on a constituency-by-constituency basis.
• All three have some claim therefore to be more trustworthy than extrapolations based on crude national swing forecasts from one party to another.

But… it’s the turnout, stupid. People – especially young people – are less likely to vote when the outcome of an election appears to be a foregone conclusion, and there’s already evidence that turnout on 4 July will be the lowest in modern British history.

In other words: as the general consensus that the Conservatives are heading for a monumental defeat grows stronger, that consensus could end up sparing the party the most catastrophic of its possible fates.

On one level this is nothing new. When turnout falls, outcomes become less clear. What makes this election different is that people are increasingly looking at polls – including the three published this week – which themselves include more detail but also more uncertainty than the old fashioned ones they’re used to.

By the numbers.

• 102 – difference between the three most recent MRP poll predictions for the number of seats the Tories will win.

• 70 per cent – share of 18-to-34 year-olds who were registered to vote in 2022.
• 355,552 – 18-to-34 year-olds who registered to vote on Tuesday, the cut-off to vote in this election.

Types. To predict general election results, pollsters have three main tools in their toolbox:

• Conventional polls. Pollsters ask roughly 2,000 people across the country who they’re going to vote for, and then publish the results. Tortoise’s average of these polls currently shows Labour winning about 41 per cent of votes nationally, the Tories winning 21 per cent and Reform winning 15 per cent. This is a measure of people, not constituencies. To predict the number of seats each party will win, the traditional approach is to find the difference between these vote shares and those of the past election and apply that swing uniformly across every constituency.

• Constituency polls. Pollsters can also do separate conventional polls in every constituency (or every constituency of interest), and then piece together the results to paint a picture for the entire country. These aren’t done much in the UK, but in the US, for example, public polling firms do conduct individual polls in important marginal congressional districts.