• New analysis by Best for Britain suggests that despite a consistent 20pt poll lead, Labour is on course secure a majority of 56-58 seats
• This landmark report on the ‘Wavering Wall’ shows undecided voters lean heavily Tory and are projected to swing more than 160 seats in a General Election
• The report has raised concerns that any narrowing of the polls over the next two years could cost Labour the next election

This landmark report into undecided voters found that despite the Labour Party’s consistent 20 point lead in the polls in the latter half of 2022, it would  deliver them a UK parliamentary majority of fewer than 60 seats.

A 10,010 person MRP poll by Focaldata taken around the resignation of Liz Truss and updated with a 2,000 person MRP poll after Rishi Sunak became Prime Minister, showed that the Labour Party’s current lead among voters would secure them more than 500 seats in a General Election, echoing the forecasts of many other recent polls.

However, many seat projections do not factor in people who respond ‘Don’t Know’ when asked how they intend to vote at the next election, despite relatively high numbers choosing this option over the past few months. In the top up poll by Focaldata, 10% of respondents said they were unsure who to vote for. 

Now, new analysis of this ‘Wavering Wall’ by internationalist campaign group Best for Britain suggests that the bulk of undecided voters are likely to be timid Tories as these respondents closely resemble the age and education profiles of people who say they intend to vote Conservative. The analysis also found that areas where there are currently higher numbers of wavering voters previously had a strong Conservative vote share and that there is a strong correlation between a reduction in the prospective Conservative voters and the growth in wavering voters between the beginning and end of 2022. 

The vast majority of these wavering voters (85%) say they would vote at the next General Election. When Best for Britain redistributed these wavering voters to other parties proportional to their age and education profiles, and taking into account the current popularity of the Westminster parties, Labour’s projected election result is slashed from an unassailable 517 seat super majority to a much more fragile 353.

With many predicting that Labour's current 20 point lead is unsustainable potentially two years out from an election, this research has raised concerns that any narrowing of the polls could cost Labour victory when Britain next goes to the ballot box.

The polls

Best for Britain commissioned Focaldata to poll a representative sample of 10,010 adults in England, Scotland and Wales with fieldwork conducted between 20-26th October 2022. We asked all respondents the following questions:

• If a general election were held tomorrow, which party would you vote for?
• If a general election were held tomorrow, how likely or unlikely would you be to actually vote?

We also asked those who answered ‘Don’t Know’ to the first questions above the following:

• Even if it is only slight, which party are you leaning towards?

Focaldata started collecting responses to the poll on 20th October which unexpectedly happened to be the day Liz Truss resigned as Prime Minister. About 75% of the fieldwork was completed before Liz Truss resigned. 

In response to this, because we could not be clear how the sudden change of Prime Minister would have affected public opinion, Focaldata ran a top-up poll of 2,000 respondents between 28-30 October 2022 once Rishi Sunak had entered Number 10. This top-up poll used the same questions as the original.

Focaldata completed an MRP model using data across all 12,000 respondents across both polls to provide constituency-level results for the question: ‘If a general election were held tomorrow, which party would you vote for?’.

Best for Britain then conducted its own analysis using data derived from the poll results and MRP constituency-level model. 

What is MRP?

Multilevel regression and poststratification (MRP) is a statistical technique for estimating public opinion in small geographic areas or sub-groups using national opinion surveys. The MRP technique has two main elements. The first is to use a survey to build a multi-level regression model that predicts opinion (or any quantity of interest) from certain variables, normally demographics. The second is to weight (post-stratify) your results by the relevant population frequency, to get population level (or constituency level) estimates. At the end of this process, you yield more accurate, more granular (thus more actionable) estimates of public opinion than traditional polling.

Focaldata’s MRP model uses a range of individual and constituency level variables, these include (but are not limited to) age, gender, education, votes at previous elections (2019 General and EU referendum), population density, % long term unemployed, % leave 2016, GE2019 vote share, Deprivation index, MP Incumbency_2018. It includes several interactions to the above as well.

All data is sourced from the Office for National Statistics (Annual Population Survey and Census) where possible, plus the Electoral Commission for election data, and estimated by Focaldata otherwise. 

Poll Results

‘If a general election were held tomorrow, which party would you vote for?’

The Conservative Party vote share increased by 6% between the two polls. Respondents answering ‘Don’t Know’ saw a 3% decrease but remained relatively high at 10%. The results for Labour, the Liberal Democrats and the SNP remained unchanged. 

It can therefore be inferred that the increased share of respondents saying they would vote Conservative in the second poll are likely to be people who would have answered ‘Don’t Know’ or ‘would not vote’ in the first. 

This report investigates the group of voters who said that they ‘Don’t Know’ which way they will vote at the next General Election. Though these voters are not a homogenous group, they do share demographic similarities. This report will refer to respondents who answered ‘Don’t Know’ in this polling as ‘wavering voters’.

Headline MRP Results

This MRP model used the combined responses from the 12,000 respondents across both polls. 

From these constituency-level results we can see that one seat (Clacton) has a higher proportion of ‘Don’t Knows’ (30%) than any party - in reality the seat would be won by Labour (29%) with a two-point lead over the Conservatives (27%).

Across the 632 parliamentary seats in England, Scotland and Wales:

• 49 constituencies had a ‘Don’t Know’ share of 25% or greater;
• 297 constituencies had a ‘Don’t Know’ share of 20% or greater;
• 527 constituencies had a ‘Don’t Know’ share of 15% or greater;
• 627 constituencies had a ‘Don’t Know’ share of 10% or greater.

The MRP model provides an average ‘Don’t Know’ result across all constituencies of 19% - meaning that based on this MRP analysis, almost one-fifth of the electorate can be classified as a wavering voter.

Analysing Wavering Voters

Those who had answered ‘Don’t know’ to the question ‘If a general election was held tomorrow, how would you vote?’ were asked: 

‘Even if it is only slight, which party are you leaning towards?’

In the most recent poll, the Conservative Party secured nearly double the support of wavering voters compared to Labour. It is reasonable to suggest that this improvement was related to Rishi Sunak becoming Prime Minister. However with a significant number of wavering voters remaining, these results were inconclusive.

For that reason, Best for Britain undertook a demographic analysis of the wavering voters, to understand whether there was evidence to support assumptions about how those voters were likely to actually cast their votes at an election. We looked at demographic data collected about the respondents to the polls, specifically age and education attainment.

See Appendix B for a full methodology for our vote transfer calculations.

Age profile of wavering voters

Best for Britain’s analysis shows there are similarities between the age profiles of those intending to vote Conservative and wavering voters. Both groups are predominantly over 55 years of age and 36% of both Conservative and wavering voters are in excess of 65 years of age. 

There are also similarities between wavering voters and those who intend to vote for Reform UK. 52% of those who intend to vote for Reform UK are aged over 55, which compares with 58% of wavering voters who are aged over 55.

Conversely, 55% of people who say they intend to vote Labour are aged 44 or below, while only 23% of wavering voters fit into this demographic.

Key Takeaway: It can be said, based on this analysis, that the age profile of wavering voters more closely resembles that of people who intend to vote Conservative or Reform UK.

Education profile of wavering voters

Similar analysis was conducted using the education attainment profile of respondents. The percentage of people who intend to vote Consevative in each education category closely mirrors the educational background of wavering voters. 

45% of Conservative voters attained a secondary school education compared to 46% of wavering voters. 24% of Conservative voters had an undergraduate or postgraduate education while 19% wavering voters did. 34% of both attended a vocational or technical college. 

Comparatively, double the number of university educated respondents say they will vote Labour (40%) in comparison to wavering voters (19%) and around quadruple the number who intend to vote Reform UK (11%).

Key Takeaway: From this analysis, it can be said that the education profile of wavering voters more closely resembles that of people who say they intend to vote Conservative or Reform UK.

Redistributing wavering voters

In Focaldata’s top up poll after Rishi Sunak became Prime Minister, wavering voters were asked how likely they were to actually vote. 85% said they either ‘Definitely would’ or ‘Probably would’ vote at the next General Election.

Therefore, to simulate a realistic General Election scenario in which wavering voters do actually vote, Best for Britain redistributed the ‘Don’t Know’ vote share to all Westminster parties to project how many parliamentary seats each party would win.

This analysis redistributed the Don’t Know share for each constituency to the other parties proportionally based on two factors: 

1. How closely the demographic profile of the wavering voters matches the profile of each parties’ prospective voters, and;
2. The vote share of each party before the redistribution of wavering voters. 

This exercise was performed twice: once using the age profile of the Don’t Knows and other party voters, and again using education profiles. Further details on the methodology can be found in the appendix of this report.

Both scenarios gave similar results, with the Labour Party securing a majority of less than 60 seats. The Conservatives lose over 130 seats compared to the results of the 2019 General Election.

In Scotland, this analysis showed the SNP had the greatest demographic and proportional similarity with the wavering voters in both scenarios and secured around 60% of wavering voters. In Wales, the Conservatives had the greatest demographic and proportional similarity with wavering voters with Plaid Cymru coming second across both scenarios.

Key Takeaway: In two different scenarios, using two different key demographic indicators, both yielded almost identical results. This supports the analysis that a significant percentage of wavering voters are more likely to vote Conservative at the next election in England and Wales, or SNP in Scotland. This greatly reduces the projected Labour UK Parliamentary majority to less than 60 seats.

The Wavering Wall - Case Studies

Below are the ten parliamentary constituencies with the highest share of wavering voters across England, Scotland and Wales. All of these constituencies are located within England, and are coincidently distributed across the country.

From the baseline MRP result, Labour are projected to gain five of the ten seats. However, when wavering voters are redistributed by both age and education profile, Labour makes no gains.

* In Clacton, Don’t Know vote share (30%) is higher than the share for any party. However, Labour is projected to win the seat with 29% ahead of the Conservatives at 27%.

One of the most extreme examples where this occurs is in Weston-Super-Mare. In the Baseline Autumn 2022 MRP, Labour are forecasted to win the seat by 11%. When redistributing wavering voters in both scenarios, the Conservatives are projected to win the constituency by a 5% margin in both scenarios, gaining 20% of wavering voters based on age profile and 19% based on education profile.

Similarly, in Clacton, Labour narrowly lead the Conservatives, but ‘Don’t Know’ has the largest share in the constituency and is forecast to win the constituency. When wavering voters in Clacton are redistributed, the Conservatives are projected to secure a 14% (Age) and 13% (Education) lead over Labour. 

Comparing Spring 2022 with Autumn 2022

To better understand the relationship between those who intend to vote for the Conservatives at the next General Election and wavering voters, Best for Britain analysed the results of the Autumn 2022 MRP alongside the results of an earlier MRP Commissioned by Best for Britain in Spring 2022.

Best for Britain compared the two MRPs to assess where these wavering voters are geographically and which party it is likely they previously intended to vote for.

Our analysis found a positive correlation between those wavering voters we identified in Autumn 2022 as wavering voters and those who said they intended to vote Conservative in Spring 2022 before the Conservatives support dropped significantly in the Summer. 

A note on the methodology for this comparison analysis and data tables can be found here.

The Spring 2022 MRP

In Spring 2022, Best for Britain commissioned Focaldata to poll the same question on Westminster Election voting intention and model the results using MRP. 10,010 UK adults were polled and the fieldwork was performed between 6 April and 29 April 2022. The results were as follows:

Which parties do wavering voters come from?

The graphs below show the relationship between the wavering voter share in each constituency (excluding Scotland) in the November MRP and the fall in each party’s vote share between May and November 2022. 

There was little correlation between a higher Don’t Know share and a fall in the Labour vote share but there was a strong positive correlation between a higher Don’t Know share and a fall in Conservative vote share. There were weak positive correlations for a higher Don’t Know share and a fall in the vote shares for both Reform UK and the Liberal Democrats.

Key Takeaway: From these results we can infer that a reduction in the Conservative vote share led to an increase in the number of wavering voters. 

Which parts of the country do wavering voters come from?

Examining the geographical distribution of wavering voters across the UK, we found that traditionally Conservative-voting areas such as Norfolk, Suffolk, the South East, and parts of the South West all have a higher wavering vote share in the Autumn 2022 Poll. 

The North East and the south of Yorkshire were exceptions to this rule. These areas exhibited a low Conservative share in May 2022, but a high ‘Don’t Know’ share in November 2022. This may be due to the greater proportion of Reform UK voters in these areas, which Best for Britain’s demographic analysis suggests is also likely to correspond with a higher proportion of wavering voters.

Key Takeaway: From these results we can see that in general, areas which previously had a strong Conservative vote share now have a higher number of wavering voters. 

Conclusion

Best for Britain’s analysis allows us to conclude that Labour’s current lead in the polls, and projected constituency-level results indicated by MRP, does not tell the full story. It appears that since Spring 2022, Conservative-leaning voters have been becoming wavering voters rather than switching to Labour.

To fully understand and predict the result of the next UK General Election, it is vital to understand who these wavering voters are and how they might cast their votes. Our polls confirm that wavering voters are overwhelmingly intending to vote, and our analysis shows they are demographically more similar to Conservative voters in England than Labour voters.

When we take into account how wavering voters are likely to actually vote, the electoral map starts to look like a closer battle for the two main parties than the headline poll results suggest.

Comparing the analyses:

Labour is currently forecast to significantly increase its share of UK parliamentary seats compared to the number it won in 2019. In each scenario outlined above Labour gains a UK parliamentary majority. 

The maps also show that areas around Yorkshire, Lancashire and the North East, sometimes referred to as the ‘red wall’, are forecast to see a significant increase in the Labour vote compared to the 2019 General Election. 

However, the table above shows that Labour’s projected majority falls by more than 160 seats when wavering voters from the Autumn 2022 MRP are redistributed based on their demographic profile, with the Conservatives forecast to gain or retain an additional 160 seats. 

Best for Britain predicts that had a UK General Election taken place in Autumn 2022 Labour would have won a clear majority, but not the majority that many would assume based on uniform national swing or by looking at MRP models alone.

Appendices and Data Tables

Appendix A: About the Autumn 2022 MRP

Appendix B: Vote transfer methodology using Age and Education profiles

Data Tables:

• Constituency-level MRP results can be found here. This includes the Autumn 2022 MRP constituency-level results as well as Best for Britain's scenarios for redistribution of ‘Don’t Know’ responses by age and education profile.

• Focaldata poll results for the 20-26 October 2022 and 28-30 October 2022 polls and Best for Britain’s calculations for redistributing wavering voters by age and education profile can be found here. 

• Best for Britain’s calculation of the ‘Don’t Know’ response share against political party can be found here.

• April 2022 MRP Constituency-level data tables can be found here.

The Report

Download the full report as a PDF here.