1 One year on
Reform would win the most seats
Current opinion polls, April and June 2025
2 The survey and the data tables
The large-sample YouGov survey
The data tables and the concepts
3 Conditional percentages
Conditional percentages … Reform, Brexit, education …
… measuring the dependence of attributes
4 Voter flow
An example: Reform voters in 2024 and in 2025
The voter flow matrix, 2024-2025
The relative attraction matrix, 2024-2025
5 Political, social, opinion
Political/social groups with the strongest support for Reform
A Britannia social cohesion scale: the Reform-LibDem differential
6 Education and experts
Brexit … educational level and experts … Michael Gove
Educational level and geographical mobility
1 One year on
“A year into power, is Starmer keeping promises?” 10
“From tears to jeers: the Labour week from hell.” 8
“Party isn’t safe on the left, even if new rival unravels.” 8
“Corbyn’s allies won’t join his ‘real alternative’.” 9
“Tories bringing back Johnson could foil Farage.” 11
“One year on. Sir Keir Starmer’s first anniversary of entering office points to a decidedly mixed record. The prime minister must ensure that the next year is better than his first.” Editorial. 29.
The Times. 5 July 2025
A year ago, Keir Starmer led the Labour party to victory in the UK general election. It was a landslide - winning 412 seats - but it was slight in terms of the percentage vote (only 33.7%). The Conservatives won just 121 seats (23.7% of the vote) and Nigel Farage’s Reform Party won a meagre 5 seats despite enjoying 14.3% of the vote.
Ten months later, this May, the Labour vote had fallen (to 23%) as had the Conservative vote (to 21%), giving Reform (at 25%) a triumph in the local elections and in a by-election.
Since then Reform (now 28%) has advanced further in the polls at the expense of the Conservatives (now 18%). So in a year, Reform has doubled its share. A recent large-sample survey by YouGov has suggested that this would translate into Reform winning 271 seats with Labour down to 178 seats, and Conservatives down to 46 seats, were an election to be held today.
Group A (Conservatives and Reform combined) had 38% a year ago and now have 46%. The Rest (Group B plus others) had 62% and now have 54%.
Reform would win the most seats
“Reform would win election held today.” The Times, June 27, 2025: 7.
The YouGov survey found that Reform would come top in an election held today, winning far more seats than Labour but well short of an absolute majority, and just short of a majority in a coalition with the Conservatives.
Table Seats in UK parliament: general election 2024; projection June 2025
. 2024 GE opinion, June 2025
Reform 5 271
Cons 121 46
LibDem 72 81
Labour 412 178
Green 4 7
SNP 9 38
Plaid 4 7
Other 23 22
Group A 126 317
Group B 488 266
Other 13+23 45+22
Total 650* 650
* includes non-voting speakers and Sinn Fein: 11=4+7?
UK 2024: https://en.wikipedia.org/wiki/2024_United_Kingdom_general_election
YouGov 2025: https://ygo-assets-websites-editorial-emea.yougov.net/documents/VotingIntention_MRP_Merge_June2025_w.pdf
Current opinion polls, April and June 2025 … July 2025
The YouGov survey discussed in the previous section had voting percentages typical of other current opinion polls: there are five parties each with less than 30% of the vote, Reform in the lead, then Labour, then Conservative, then Liberal Democrat and then Green. Group A and Group B are equal.
April: “The median of five recent opinion polls put the Conservatives on 21%, ranging between 20% and 23%. Labour and Reform have similar medians, 23% and 25% respectively. Lower are the Liberal Democrats with a median of 14%; and the Greens with a median of 9%.”
June: Since April, the Conservatives have fallen further (18%) with a corresponding rise in Reform (28%)
July: the first July opinion poll gives similar results
Grouping the parties, Group A and Group B are equal, at 47%. Group A has Conservatives and Reform. Group B has Labour, Lib Dem and Green. (In other reports I have included SNP in Group B.)
A major topic of discussion is a possible pact or understanding between the two Group A parties, Conservative and Reform.
. April … June
A Conservative 21% (20-23) … 18% (17-20)
B Labour 23% (20-26) … 23% (22-27)
A Reform 25% (25-28) … 28% (27-30)
B Lib Dem 14% (11-16) … 14% (12-16)
B Green 9% ( 8-13) … 9% (7-11)
…
Group A 47% (45-48) … 47% (44-49)
Group B 47% (45-49) … 46% (46-49)
…
Source
UK opinion polls, end of polling date: 21-25 April 2025; 23-26 June 2025, five polls
https://en.wikipedia.org/wiki/Opinion_polling_for_the_next_United_Kingdom_general_election
Postscript
John Curtice discusses the latest poll by More in Common.
“Party isn’t safe on the left, even if new rival unravels.” The Times, 5 July 2025. 8
New Statesman: https://www.newstatesman.com/politics/polling/2025/06/the-phantom-threat-of-corbyn-2-0
2 The survey and the data tables
The large-sample YouGov survey
The large-sample YouGov survey involved 17,227 people. This sample is much larger than the regular surveys of perhaps a couple of thousand people conducted by YouGov and other opinion pollsters. In one sense a large sample allows estimates to be much more accurate (but it does not guarantee an unbiased sample).
YouGov used the large sample to produce estimates of voting at the constituency level using MRP methodology and hence to predict the number of seats that each party would gain were a general election held today (as discussed above).
https://yougov.co.uk/politics/articles/52437-first-yougov-mrp-since-2024-election-shows-a-hung-parliament-with-reform-uk-as-largest-party (11,500 voters)
Also available are the tables of results (1) which show the overall voting percentages and also the percentages broken down by a variety of political and social characteristics. The link to (1) was obtained from the Wikipedia page on UK opinion polls (2).
.(1) “YouGov Survey Results” 2025: https://ygo-assets-websites-editorial-emea.yougov.net/documents/VotingIntention_MRP_Merge_June2025_w.pdf (17,227 voters)
. (2) UK opinion polling:
https://en.wikipedia.org/wiki/Opinion_polling_for_the_next_United_Kingdom_general_election
The data tables and the concepts
The “YouGov Survey Results” gives the sample size as 17227 adults in GB, noting that the fieldwork was carried out between 29 May and 18 June 2025. In the tables, all the percentages are the voting percentages for the parties based on the responses by the people in the sample. The first vertical column gives the overall voting percentages for the parties. The other vertical columns give the subgroup voting percentages for the parties. One might refer to the latter percentages as conditional percentages, that is percentages conditional on the subgroup. In each vertical column the percentages add up to 100%.
The heading for each table gives the different subgroup categories and the subgroups within each category. For each subgroup it gives the number of people in that subgroup, both the number in the unweighted sample and the number in the weighted sample.
On each page there are two tables. The second table gives the results for all the sample, the first table gives the results for all except those who did not give a specific response to the main question.
The first table is:
“HEADLINE VOTING INTENTION
Westminster voting intention
[Headline voting intention projected by the MRP model]”
The second table notes the question:
“Now, thinking specifically about your own constituency, if there were a general election held tomorrow and these were the parties standing, which party would you vote for?”
It includes a warning:
“*Any percentages calculated on bases fewer than 100 respondents do not represent a wide enough cross-section of the target population to be considered statistically reliable. These figures should not be used.”
It notes:
YouGov weights GB political surveys by (1) age interlocked with gender and education, (2) political attention (3) social grade (4) 2024 recalled vote interlocked with region (5) EU referendum recalled vote and (6) ethnicity. Weighting targets are YouGov estimates, data sources used in the calculation of targets are cited below. Voting intention figures are separately modelled using MRP. The poll was carried out online. YouGov Weighting Data All polls are subject to a wide range of potential sources of error. On the basis of the historical record of the polls at recent general elections, there is a 9 in 10 chance that the true value of a party’s support lies within 4 points of the estimates provided by this poll, and a 2 in 3 chance that they lie within 2 points.
For more information on the methodology, please go to https://yougov.co.uk/about/panel-methodology
As noted above the tables give what might be called the ‘vertical conditional percentages’: they add up to 100% when added vertically. What the tables don’t give directly are the ‘horizontal conditional percentages’. However these can be calculated from the numbers in the table. The following section gives a more detailed account of conditional percentages.
3 Conditional percentages
Conditional percentages … Reform, Brexit, education
Given the dramatic rise in Reform, it is of interest to identify the variables which are related to the Reform vote. This section considers the relationship with three variables: previous Reform voting, Brexit voting and educational qualifications. The relationship is studied using the concepts of conditional percentages and dependence.
What percentage of Reform voters then are Reform voters now? 93%
… loyalty
What percentage of Reform voters now were Reform voters then? 40%
… previous voters have been added to by new recruits
[then, July 2024; now, June 2025]
What percentage of Leave voters then are Reform voters now? 51%
… half take this path, half have moved elsewhere
What percentage of Reform voters now were Leave voters then? 94%
… virtually the only source
[then, 2016; now, June 2025]
What percentage of people with no educational qualifications are Reform voters? 50%
… half the people with this attribute are Reform people
What percentage of Reform voters have no educational qualifications? 9%
… a rare attribute of Reform people
Note 1. The notion of conditional percentages is akin to that of conditional probabilities and less directly to the distinction between the regression of X on Y and the regression of Y on X.
Note 2. The above percentages 93%, 51% and 50% can be found in the tables given in the link below. The tables give what might be called the ‘vertical conditional percentages’: they add up to 100% when added vertically.
What the tables don’t give directly are the ‘horizontal conditional percentages’ (here 40%, 94% and 9%). However they can be calculated from the numbers in the table.
… measuring the dependence of attributes
This section discusses the algebraic equations for conditional percentages and dependence.
Probability theory gives the condition when two events are independent of one another: this is the case if the probability of both events equals the product of the probabilities of the two events. From this we can define a simple measure D of dependence. If D=1 then the two events are independent.
In general, given two sets G and H, there is a Venn Diagram:
those in both G and H;
those in G but not in H;
those in H but not in G;
those in neither.
These correspond to the subsets G&H, G-H, H-G, and notG¬H.
In general, given two groups G and H, there are two conditional percentages:
What percentage of group G are in group H?
What percentage of group H are in group G?
Percentages are denoted as P(G) and P(H); and P(G&H), P(G-H), P(H-G), and P(notG¬H); and conditional percentages as P(G/H) and P(H/G).
The conditional percentages are given by:
P(G/H) = P(G&H) / P(H)
P(H/G) = P(G&H) / P(G)
This can be rewritten as:
P(G&H) = P(G/H) P(H) = P(H/G) P(G) [1]
A special case, the case of G and H being independent of one another is if
P(G&H) = P(G) P(H) [2]
In this case of independence:
P(G/H) = P(G)
P(H/G) = P(H)
A measure of dependence D is given by:
D = P(G&H) / P(G) P(H) [3]
= P(G/H) / P(G) [4]
= P(H/G) / P(H) [5]
From [3] we have
P(G&H) = D P(G) P(H) [6]
In the special case of [2], equation [3] gives D=1.
Equation [6] can be expressed as:
joint percentage equals dependence times basic percentage times basic percentage
… and in terms of flows:
flow rate* equals relative attraction** times first percentage times second percentage
* from first to second; ** of second for first
4 Voter flow
An example: Reform voters in 2024 and in 2025
Some people voted Reform in 2024 and some people voted Reform in 2025.This can be thought of in terms of the following concepts. In terms of the next two sections, item (4) relates to the flow rate (in this example 93%); and item (7) relates to the relative attraction (in this example 3.44).
Basic percentages
.(1) The percentage voting Reform in 2024: 11.5%. (in the YouGov sample)
.(2) The percentage voting Reform in 2025: 27%.
Conditional percentages
.(3) For those voting Reform in 2024 then, the percentage voting Reform in 2025 now: 93%.
.(4) For those voting Reform in 2025 now, the percentage voting Reform in 2024 then: 40%.
Joint percentage (the percentage voting Reform in both 2024 and 2025)
.(5) If independent then, multiplying (1) and (2): 3.1%.
.(6) But actually not independent: 10.7%.
A measure of dependence D … ‘relative attraction’ A
.(7) Dividing (6) by (5) … or dividing (3) by (2) … or dividing (4) by (1) …
… D = 3.44
The voter flow matrix, 2024-2025
In the previous section we found that 93% of those voting Reform in 2024 went on to vote Reform in the YouGov poll in 2025. We refer to this as the staying rate for Reform. The table below shows this and the staying rates for the other parties: 62% for Conservatives, 74% for the Liberal Democrats, 83% for Labour and 77% for Greens.
The table also shows the flow rates from one party to another. There are strong flows from Conservatives to Reform (30%); and from Labour to Liberal Democrat (12%) and to Green (10%). These strong flows are respectively within Group A and within Group B.
In general there is some flow of voters between Group A and Group B, mostly from B to A, but there is an around 90% staying rate within each group. LibDem flows almost as much to Group A as within Group B, whereas Labour and Green flow mainly within Group B.
The category ‘Other 2024’ includes SNP and Plaid Cymru. This category had a staying rate of 71%. In particular the category had 42% for SNP in 2025. The category divided equally between Groups A and B, for 2025 – 16% and 15%. Reform attracts 13%, Labour only 6%.
The category “dn” includes those who did not vote in 2024 and those who did not know (?). The category split almost equally in 2025: 49% for A, 42% for B and 8% for Other. Reform attracts 35%, Labour only 20%.
Table The voter flow matrix, 2024-2025
.
. Group A, 2024 Group B, 2024 …………….
. R C dn24 LD L G Other 2024
Group A
Refor 93* 30 35^ 6 8 3 13
Cons 3 62* 14 4 3 3 3
Group B
LibD 1 4 12 74* 12 7 3
Lab 1 2 20^ 8 63* 7 6
Green 1 1 10 6 10 77* 6
Other 2025
SNP 0 0 4 0 1 1 42
Plaid 0 0 1 0 1 1 11
Other 2 1 3 1 1 2 18
Grp A 96* 92* 49* 10 11 6 16
Grp B 3 7 42* 88* 85* 91* 15
Other 2 1 8 1 3 4 71*
Not included in the above table are those who, in the YouGov poll 2025, in an expanded table with an expanded data base, would not vote (12%); don’t know (13%); or refused (3%), labelled ‘dn25’ below. The staying rate for this absence of specific response group is 60%. Note that in this expanded data base between 9% and 24% of those who voted for some specific party in 2024 did not give a specific preference in the 2025 YouGov poll.
Table The voter flow for dn25, expanded data base
. R C dn24 LD L G Other 2024
dn25 9 14 60* 15 19 15 24
The relative attraction matrix, 2024-2025
The previous section shows the flow rates, which are simply the vertical conditional percentages given in the YouGov tables. The rates do not tell us whether or not there is dependence between the percentages in 2024 and the percentages in 2025, nor do they give a measure of such dependence. The table below gives these measures of dependence. In the context of flows, the measure of dependence is referred to as the relative attraction
There is some similarity between the two tables. There is a correspondence in the asterisked cells in the two tables, reflecting the fact that high staying rates (flow rates) correspond to high positive dependence (relative attraction). In the case of Reform the high flow rate of 93% corresponds to the high dependence (relative attraction) of 3.44.
The ordering of the parties in the table – R, C, LD, L, G - is such that each row and column exhibits a single-peaked profile – with just a few exceptions. In other words relative attraction is at a peak for one’s own party and decreases the farther away one moves from one’s own party along the dimension.
For example the Conservatives in 2024 have a peak relative attraction of 3.65 for themselves in 2025, followed by 1.11 for Reform, 0.27 for LibDem, 0.9 for Labour, 0.9 for Green, 0.5 for Other and 0 for SNP and Plaid Cymru in 2025.
Table The relative attraction matrix, 2024-2025
.
. Group A, 2024 Group B, 2024 …………….
. R C dn24 LD L G Other 2024
Group A
Refor 3.44* 1.11 1.30 0.22 0.30 0.11 0.48
Cons 0.18 3.65* 0.82 0.24 0.18 0.18 0.18
Group B
LibD 0.07 0.27 0.80 4.93* 0.80 0.47 0.20
Lab 0.04 0.09 0.87 0.35 2.74* 0.30 0.26
Green 0.09 0.09 0.91 0.55 0.91 7.00* 0.55
Other 2025
SNP 0 0 1.33 0 0.33 0.33 14.00*
Plaid 0 0 1 0 1 1 11.00*
Other 1 0.5 1.5 0.5 0.5 1 9.00*
Grp A 2.18* 2.09* 1.11 0.23 0.25 0.14 0.36
Grp B 0.06 0.14 0.86 1.80* 1.73* 1.86* 0.31
Other 0.29 0.14 1.14 0.14 0.43 0.57 10.14*
In terms of an earlier section, from [3] we have
P(G&H) = D P(G) P(H) [6]
Elsewhere I have discussed this in terms of a gravity model, giving essentially the same equation where F is flow, a is the “relative attraction” and the p’s are the two basic percentages – the relative attraction being the specific attraction divided by the general attraction
F = a p p’
5 Political, social, opinion
Political/social groups vary in the percentage of members who support Reform.
The groups which have the strongest support for Reform are those who voted Reform in 2024 (93%, see earlier section), those who voted leave in 2016 (51%) and those with no or only GCSE educational qualifications (50% and 41%). Otherwise groups with low political attention, older, retired, lower social grade, LA renting, owning home outright, lower household income ,,, have between 30% and 39% of their members supporting Reform. It is more precise to think of a correlation between group membership and support for Reform rather than absolute statements that members of those groups support Reform.
It may be that opinions correlate more strongly with Reform than does membership of political/social groups. Thus a previous report proposes a Britannia social cohesion scale: the Reform-LibDem differential.
Reform might be seen as having a self-positive other-negative view, with negative others including the European Union, immigration and multi-ethnic society.
Political/social groups with the strongest support for Reform
This section continues the discussion of conditional percentages, now looking at a wider range of variables. The first table identifies the groups which have the strongest relationships with Reform, leaving the second table to give the detailed results for the four main parties.
Groups vary in the percentage of members who support Reform. Here are the groups which have strongest support for Reform (in terms of voting intention today).
Table The support for Reform in various political/social groups
Reform voters today constitute …
political
93% of those who voted Reform in 2024
51% of those who voted Leave in 2016
34% of those with low political attention
education
50% of those who have no educational qualifications
41% of those who have GCSEs but nothing higher
[also other or don’t know answers to education question]
age, work, class
36% of those who are over 65 years old
36% of those who are retired
[also other answers to work question]
35% of those in social classes C2, D, E
housing
39% of those who rent from LA or HA
32% of those who own outright
household income
32% of those who have less than £20,000
31% of those who have £20,000 to £30,000
31% of males
27% of whole sample
The table below gives the support for four parties in various political/social groups.
Table The support for four parties in various political/social groups
. L: Labour; LD, LibDem; C, Conservative; R Reform
. L LD C R
Brexit
Rem 35 22 14 8
Leave 9 8 25 51*
education
none 12 10 18 50*
other 15 13 21 42*
GCSE 16 10 20 41*
A level 22 13 17 29
<degr 25 15 17 26
degree 32 19 14 13
dk 17 9 12 41*
ot tech 19 16 25 28
age
18.24 33 16 7 11
25.48 30 16 12 21
50.64 21 15 17 34*
65+ 13 14 30 36*
class
AB 29 17 19 18
C1 24 15 18 24
C2 19 14 15 35*
DE 19 12 17 35*
work
work 27 16 15 24
stud 33 18 9 7
retire 14 13 28 36*
unemp 26 12 8 24
notw 23 12 14 29
other 17 18 8 37*
Housing
own 18 15 25 32*
mortg 27 16 16 23
rent la 21 10 11 39*
rent pr 28 16 11 21
neither 36 18 11 19
political attention
low 18 12 18 34*
med 23 15 19 27
high 29 16 14 23
household income £,000
<£20 19 15 17 32*
20-30 21 13 17 31*
30-50 24 13 19 26
50-70 29 15 14 24
>70 31 19 17 17
gender
male 26 14 16 31*
femal 22 16 19 24
A Britannia social cohesion scale: the Reform-LibDem differential
“The report gives a breakdown by political party for seven of the questions. For each question the net percentage for each political party can be calculated. On the basis of these percentages the political parties can be ordered as follows – Reform, Conservative, Labour, LibDem and Green – this ordering giving either increasing or decreasing net percentages for each question (with some exceptions). For each question the Reform-LibDem differential can be calculated.
The multiculturalism question has the highest differential at 95% with Reform much more likely to agree that multiculturalism threatens our national identity. The score of 95% involves the difference between the 60% net Reform voters agreeing and the 35% net LibDem voters disagreeing.
The next highest differential related to feeling a stranger in their own country (84), followed by identification of the Southport rioters (68) and integration being the responsibility of ethnic minority groups (62).”
See the final section in:
Reform and Britannia … Poland … Love All Always or Prefer Those Like Ourselves?
6 Education and experts
An association has been noted above between education and support for Reform. This association is referred to in the remarks below by Michael Gove. A recent report discusses how education relates to social and geographical mobility, processes which perhaps help create support for Reform.
Brexit … educational level and experts … Michael Gove
“Some of the things which made me more pro-Brexit were the reactions to it. How the condescension towards people beforehand became more vivid and strident afterwards. ‘these people didn’t know what they were voting for’, or, ‘voting Leave is correlated with a lower level of education.’ That’s only made me more pleased to have been on that side.”
[Note that Remainers also can have a self-positive other-negative view.]
“… people have had enough of experts … [by that I mean] people in organisations with three-letter acronyms. The IFS, CBI, IFP. And so on. Book-smart people who attempt to reduce the complexity of humanity to something that will fit into a PPE essay.”
“How not to be a political husband.” The Times, Magazine, 23 June 2025: 16-21.
Educational level and geographical mobility
“Clever teenagers have left their home town by age 32.” The Times, 23 June 2025: 2.
Xiaowei Xu, IFS.
https://ifs.org.uk/news/six-ten-top-achievers-outside-london-leave-their-home-town-their-early-30s;
“Geographical inequalities in the UK are large and persistent. Previous research has found that most of the gap in earnings across places – the main driver of local living standards – reflects differences in the kinds of people who live in different places (Overman and Xu, 2024). At the same time, young people in England are highly mobile, especially at the start of their working lives (Britton et al., 2021). Understanding young people’s mobility – who moves, where and why – is crucial to understanding how geographical inequalities arise, and why they persist.
We consider disparities across local labour markets in Great Britain. Disparities in wages and employment rates are large and persistent, although smaller than 20 years ago. These disparities largely reflect the concentration of high-skilled workers, who would have better labour market outcomes wherever they live. This concentration is driven by differences in the demand for, and supply of, skills and the self-reinforcing interaction between the two, which is particularly pronounced in the highest-wage areas and at the upper end of the wage distribution. The highest-paid jobs are concentrated in London and a handful of other areas and wage disparities are mostly driven by the higher-paid. Places that offer higher earnings also have higher rents, which may entirely offset gains in earnings. Consistent with this, people in higher-paid places are no happier than those in lower-paid places.
https://academic.oup.com/ooec/article/3/Supplement_1/i585/7708135?login=false
In this report, we study the link between higher education, geographical mobility and outcomes for individuals and regions. First, we examine whether higher education is associated with greater geographical mobility even when we control for confounding factors, such as the fact that graduates tend to be wealthier and have higher educational attainment. We examine how mobility – and the relationship between higher education and mobility – differs across socio-economic and ethnic groups. Second, we consider whether graduates do indeed seem to be ‘moving to opportunity’ by analysing the types of areas graduates move to and from and the earnings gains associated with moving. Third, we consider the effect of mobility patterns on regional inequality.”
THE END