Making sense of the polls

In the run-up to the general election there are polls a’plenty. But what do they actually tell us? And can we trust them? Nick Terdre, text, and Russell Hall, research and graphics, dig into the current crop and what they tell us about the likely outcome of the general election on 4 July.

The YouGov poll published in mid-week puts Labour’s vote share at 36% and the Conservatives at 20%, followed by Reform UK 18%, the Liberal Democrats 14%, the Greens 7%, the Scottish National Party 3%, Plaid Cymru 1% and others 2%.

The polling company’s projected division of seats is Labour 425, Conservatives 108, LibDems 67, SNP 20, Reform five, Plaid Cymru four and Greens two, while the Speaker’s seat, currently in Chorley, is traditionally not contested. As with the other polls discussed here, these results apply to the 632 parliamentary seats on the UK mainland, and excludes the 18 seats in Northern Ireland.

So, thanks not least to the skewing effects of our first-past-the-post electoral system, we seem to be heading for a Labour landslide, with a likely majority of 200 seats, and according to some media speculation, the possible extinction of the Tory party.

But should we get carried away by such a cataclysmic prospect? Can we trust the polls? What do they actually tell us?

There have, after all, been some spectacular failures by polls to forecast the outcome of general elections in the not-so-distant past. In 1992 a Labour victory was confidently forecast but a further five years of Tory rule ensued. In 2015 the Tories and Labour were neck-and-neck, according to the polls, but the Tories ended up with a solid majority.

After the 2015 election the British Polling Council, an association of polling companies whose main aim is to encourage high standards of disclosure of survey results, was highly critical of the way the polls had been conducted, and called for changes. These were made and the results were better in 2017 and 2019.

Nowcasts or forecasts?

Nevertheless, all polls should be treated with some scepticism. The question asked by the YouGov pollsters was: If there were a general election tomorrow, which party would you vote for?

So what the results reflect is how respondents intend to vote at the time they are asked — in other words, they are ‘nowcasts’ rather than ‘forecasts.’ And between now and election day, minds may change, even though there is less than two weeks to go.

The results may also be skewed, for example by the ‘shy’ voter, who conceals their true voting intention. Given the Conservative party’s current travails, some of their supporters are thought to be reluctant to reveal they will vote for it, which could lead to an underestimate of the party’s vote share.

Are polls reliable? Can we really extrapolate from the data generated by a limited set of questions asked of a limited sample of respondents to form an accurate picture of how a large population intend to vote, especially at constituency level? Even a relatively large sample of 50,000 is equivalent to an average of only 79 people in each constituency.

It’s a perennial question for pollsters, who on top of stratifying their sample to match as closely as possible the known demographics, have developed sophisticated statistical processes for analysing their raw data.

Currently in mode is the multi-level regression and post-stratification technique, MRP, which was first used for a UK general election in 2017. This enables more nuanced information to be teased out at constituency level. All the polls quoted in this article make use of the MRP method.

And while it is the headline figure which grabs the attention, pollsters are careful to present their projections as a range — normally there is a 90% probability that the true value lies within the range.

It is also worth pointing out that according to the YouGov poll there are 109 so-called toss-up seats where the current leader is 5% or less ahead, so it would not take much of a swing to produce a different result.

As the table shows, there are 46 seats in which the Conservatives have a lead of 5% or less and are vulnerable to being overtaken by the chasing party, which in 33 of them is Labour. Similarly in 34 of the 40 seats in which Labour has only a small lead, it is the Conservatives who are breathing down their neck.

What can we learn from the polls?

So bearing all these provisos in mind, what do current polls tell us about the likely outcome of the general election less than two weeks away?

Certainly there appears to be a dramatic change in prospect in the division of seats from the Parliament which has just been dissolved, which was Conservatives 345, Labour 206, SNP 43 and LibDems 15.

Looking at the table of MRP polls carried out at various times in the period since early April, both before and after Prime Minister Rishi Sunak announced the election date, Labour appears to be heading for a substantial victory, with a majority estimated between 114 and 382.

The calculation of Labour's majority takes into account the 18 Northern Ireland seats and any seats won by parties or independents not listed in the table.

The stand-out event to have shifted the dial in voting intentions is Nigel Farage’s announcement on 3 June that he would take over as Reform UK’s party leader and attempt to break his long-standing parliamentary duck – seven failures so far – by standing in Clacton.

Reform’s improved fortunes since then are a good example of how quickly things can change. Prior to Farage’s intervention it was running third, well behind the Conservatives, and not expected to win any seats. But in an update on 13 June YouGov put it marginally ahead of the Tories in voting intentions for the first time ever, 19% against 18%.

Reform could find itself with a presence in the next Parliament — at best the MRP polls see them winning seven seats, but maybe none.

Where the polls differ quite substantially is how the seats will be divided up. Labour is seen taking between 382 and 516 seats, the Tories between 53 and 180, the LibDems between 30 and 67; the latter would represent a record number for them, and they anyway look likely to overtake the SNP as the third largest party.

Interactive map showing projected results for each constituency based on YouGov's second MRP poll.

Local level

Now let’s look at the local level, using seat projections from Survation based on fieldwork done between 31 May and 13 June, including a period after Farage’s announcement. In Hastings and Rye, Survation puts Labour candidate Helena Dollimore on 56% and the Conservatives’ Sally-Ann Hart on 25.1%. Even the top end of the 90% credible range for Hart, 31.9%, is well below the lower end of the range for Dollimore, 47.1%.

Reform’s Lucian Fernando has 12%, and an upper range of 20.5%. He seems to have benefited from the Farage bounce — polls based on fieldwork carried out before Farage’s intervention gave him at best 10%. However, he has by no means caught up with Hart.

The Survation poll projects a vote share of 3.7% for the LibDems’ Guy Harris, 1.4% for the Greens’ Becca Horn and 0.3% for the others — Nicholas Davies of the Communist Party of Britain, Phil Colley of the Workers’ Party of Britain, and independent Paul Crosland.

YouGov also sees Labour with a big, though lesser, lead over the Tories — 51% against 26%, making it a likely Labour gain. However, compared with its earlier MRP poll, Labour's vote share is down by 9% and the Tories' by 3%, while Reform is up by 7% and the Green Party by 6%, as the chart below shows.

Rather like Sunak, Hart has attempted to bolster her position by moving further to the right — last month she announced that she was in favour of the UK withdrawing from the European Court of Human Rights (a measure Sunak declined to include in the Tory manifesto).

Incidentally, one surprise supporter Hart has attracted is Claire Carr, ex-local councillor for Labour and then the Greens, who has been spotted in her canvassing entourage and is one of the assentors to her candidacy.

Close race in Bexhill & Battle

Meanwhile, a closer race is developing in Bexhill and Battle, where the Tories have selected Kieran Mullan, former MP for far-off Crewe and Nantwich to replace Huw Merriman, one of 75 Tory MPs who have decided to stand down (not all, of course, because they don’t have the stomach for a fight).

Bexhill and Battle appears to be swinging towards Labour. While YouGov’s previous poll saw it as a Conservative hold, 38.3% against 32.1% for Labour’s Christine Bayliss (of Rother District Council), its latest survey puts it in the toss-up category, with Labour at 33.5% and the Tories at 30.5%.

Survation’s latest poll gives Labour a bigger lead – 34.2% against 28.8% – though as the graphic below shows, there is considerable overlap between the ranges. The upper end of Reform's range is well within Labour's range. It is also notable that while both the Labour and Tory vote share has gone down since the company’s previous survey, by 2.8% and 2.4% respectively, Reform’s Ian Gribbin is up by 4.6% to 19.9%.

The assumption of tactical voting is baked into most polls, including YouGov, though Electoral Calculus produces two sets of projections, with and without tactical voting.

This could make a difference in Lewes, according to the pollster, whose projection without tactical voting is a Tory hold — 31.8% for the incumbent Maria Caulfield against 27.9% for the LibDems’ James MacCleary. But if Labour and Green supporters lend some of their votes to the LibDems, that could lift them to 34.8%, above the Tories’ 33.1%.

Meanwhile, YouGov sees it as a safe LibDem gain — 51% against 22.3%.

In Eastbourne most polls, including the latest YouGov, project a LibDem gain and a Tory hold in the Sussex Weald (Wealden following boundary changes).

We’ll have to wait until after 4 July to find out just how accurately the polls managed to capture voters’ intentions and project the destiny of individual constituencies.

But even armed with sophisticated statistical techniques their task is not an easy one. Voter behaviour is changing and becoming less predictable. Voters are more likely to change party — as witnessed by the fall of the “red wall” seats to the Tories in 2019. Tactical voting – not least impelled by the unfairness of the first-past-the-post system – is also on the increase, adding to voter unpredictability.

As well as shy respondents, there are also undecided ones, and different pollsters treat them differently. YouGov polling on 12-13 June found 18% of 2019 Tory voters still undecided, as well as 11% of LibDem voters and 8% of Labour voters. As the election gets nearer, more will make up their minds, so later polls should in this respect at least produce more reliable results.

The boundary changes that have been made since 2019 also complicate life for the pollsters, who draw on the results of the previous general election in their modelling.

On the other hand, since 7 June, when nominations closed, voters have known who the candidates are, which helps them to make up their minds.

So to sum up: there is much to bear in mind when reading the latest poll results, but the overriding principle is — nothing is cast in stone!

Also in: Elections

« Hustings: more tickets available
HOT & HIP Hustings: Thursday 27 June »