Brexit vote explained: poverty, low skills and lack of opportunities

31st Aug 2016

This report provides unprecedented insight into the dynamics of the 2016 vote to leave the EU, showing how a lack of opportunity across the country led to Brexit.

Devoting specific attention to data on the roles of poverty, place and individual characteristics driving the leave vote, this report shows how Britain was divided along economic, educational and social lines.

In the aftermath of the vote few studies have considered both individual and area-level drivers of the vote to leave the EU. This report reviews existing research, examines new data and considers implications for the wider debate.

Key findings:

  • The poorest households, with incomes of less than £20,000 per year, were much more likely to support leaving the EU than the wealthiest households, as were the unemployed, people in low-skilled and manual occupations, people who feel that their financial situation has worsened, and those with no qualifications.
  • Groups vulnerable to poverty were more likely to support Brexit. Age, income and education matter, though it is educational inequality that was the strongest driver. Other things being equal, support for leave was 30 percentage points higher among those with GCSE qualifications or below than it was for people with a degree. In contrast, support for leave was just 10 points higher among those on less than £20,000 per year than it was among those with incomes of more than £60,000 per year, and 20 points higher among those aged 65 than those aged 25.
  • Support for Brexit varied not only between individuals but also between areas. People with all levels of qualifications were more likely to vote leave in low-skill areas compared with high-skill areas. However, this effect was stronger for the more highly qualified. In low-skilled communities the difference in support for leave between graduates and those with GCSEs was 20 points. In high-skilled communities it was over 40 points.  In low-skill areas the proportion of A-level holders voting leave was closer to that of people with low-skills. In high-skill areas their vote was much more similar to graduates.
  • Groups in Britain who have been ‘left behind’ by rapid economic change and feel cut adrift from the mainstream consensus were the most likely to support Brexit. These voters face a ‘double whammy’. While their lack of qualifications put them at a significant disadvantage in the modern economy, they are also being further marginalised in society by the lack of opportunities that faced in their low-skilled communities. This will make it extremely difficult for the left behind to adapt and prosper in future.

Introduction

The 2016 vote to leave the EU marked a watershed moment in the history of the United Kingdom. When all votes had been counted 52% of the electorate had opted to leave the EU, a figure that increased to almost 54% in England. The figures for Scotland, Wales and Northern Ireland varied, at 38, 52.5 and 44%, respectively. Across 393 local authorities, support for leave surpassed 70% in eight authorities, 60% in 102 and 50% in 263. In England the share of the vote for leaving ranged from nearly 76% in Boston, Lincolnshire to 21% in Lambeth, London.

Like Boston, many local authorities that recorded some of the strongest support for Brexit are struggling areas where average incomes, education and skill levels are low and there are few opportunities to get ahead. Authorities that recorded some of the highest levels of support for Brexit include the working-class communities of Castle Point, Great Yarmouth, Mansfield, Ashfield, Stoke-on-Trent, and Doncaster. In such communities the types of opportunities and life experiences contrast sharply with those in areas that are filled with more affluent, highly-educated, and diverse populations, which gave some of the strongest support to remaining in the EU, such as Islington, Edinburgh, Cambridge, Oxford and Richmond upon Thames.

The geography of the vote has sparked a debate about a divided Britain in which many have traced the vote for Brexit to economically disadvantaged and low skilled ‘left behind’ communities that amid a post-industrial and increasingly global economy are struggling to keep pace with high skilled areas. But to what extent is this interpretation supported by data? What motivated the vote to leave the EU and what role did poverty and place play in these decisions?

Our aims are two-fold. First, building on work by the Joseph Rowntree Foundation (JRF) we examine the relationship between poverty and the vote for Brexit. One of the most contested issues in the referendum campaign was the claimed economic impact of Brexit. This debate is linked closely to poverty, since the question of the short- and long-term effects of Brexit on the economy – including on exchange rates, growth, investment and employment – all have direct and indirect effects on the poorest people and places, as also noted by JRF.

Between 2011 and 2014, nearly one-third of the UK population experienced relative income poverty at least once. Groups most vulnerable to poverty are older people, people who left school without any formal education, women, and people in single-person households. The chances of entering poverty also vary across different areas. Whereas some areas are thriving, others are in decline. A recent report by JRF shows that this decline consists of numerous factors such as population loss, those with higher skills moving out, economic restructuring and de-industrialisation, shrinking labour markets, unemployment, low education and skills, poor health, deprivation and poverty, physical blight and declining tax bases. But were poverty and place central drivers of the vote to leave the EU? To explore this question, we have undertaken new research to offer hitherto unprecedented insight into the dynamics of the vote.   

Second, we present findings from new research on individual voters who readily identified themselves as supporters of Brexit. Until now, much of the research on the referendum has focused on the area or ‘aggregate’ level, exploring for instance the relationship between the characteristics of communities and their levels of support for leaving or remaining in the EU. But looking only at the area level masks what is happening at the individual level. For example, knowing that lots of Eurosceptic voters live in Clacton is helpful but it does not really tell us much about why those individuals in Clacton actually decided to vote for Brexit. In this report we push the debate forward by considering both the area and individual-level drivers of support for Brexit as well as how these interact. Drawing on data from the British Election Study (BES), we put the backgrounds, attitudes and values of leave voters under the microscope, painting a detailed picture of what motivated their decision at the referendum. This allows us to contribute to the national debate, exploring what the findings reveal about issues that need addressing in relation to poverty, skills and opportunity, and in different parts of the country.

Voting to leave the EU: existing research

The relationship between poverty, education and age, and level of support for Brexit

A first step to examining the role of poverty in the referendum vote is to examine existing work on the relationship between the characteristics of areas and their level of support for Brexit (or what academics call the ‘aggregate level’). Broadly speaking, past research traces support for Brexit to areas with older populations and lower than average levels of education. These areas are more likely than others to experience deprivation and, in recent years, witnessed significant demographic change as a result of the inward migration of EU nationals.  

In the immediate aftermath of the referendum our earlier work (Goodwin and Heath, forthcoming, see Reference notes below) examined data from 380 of the 382 local authorities across the UK, linking this to information from the 2011 census. We found that support for Brexit was strongest in areas where a large percentage of the population did not have any qualifications and were ill-equipped to thrive amid a post-industrial and increasingly competitive economy that favours those with skills and is operating in the broader context of globalisation. For instance, 15 of the 20 ‘least educated’ areas voted to leave while all of the 20 ‘most highly educated’ areas voted to remain. Support for Brexit was also stronger than average in areas with a larger number of pensioners. Of the 20 youngest authorities 16 voted to remain, but of the 20 oldest authorities 19 voted to leave.

Such findings are generally consistent with past work on support for UKIP that talked about the importance of ‘left behind’ communities. Examining areas where support for UKIP was strongest, Matthew J Goodwin and Caitlin Milazzo (see also Robert Ford and Matthew J Goodwin 2014) noted how this was often ‘in areas where there were lots of older, white and poor voters, while it was consistently weaker in areas that were younger, more ethnically and culturally diverse, and financially secure’. It was these differences in local demography that helped to explain why UKIP won 40% of the vote in economically struggling places like Rotherham but only 14% in the more affluent and leafy Richmond upon Thames.

The relationship between deprivation and level of support for Brexit

However, others warn against an interpretation of the vote that focuses only on economic insecurity. One early analysis of the referendum result by Alisdair Rae suggests that while there is a strong correlation between support for Brexit and the percentage of people who have no qualifications this support was not strongly correlated with deprivation. Yet such findings stand at odds with other work. The Financial Times commissioned two economists to shed light on the relationship between wage growth and – as a proxy for the leave vote –past support for the UK Independence Party (UKIP). They found a statistically significant link between a lack of wage growth and the share of the vote going to UKIP at the 2015 general election. In working-class and struggling communities like Castle Point in Essex the real median wage had declined by 13% since 1997. Based on these findings Sarah Neville suggested that the gloomy economic forecasts released by the remain campaign had failed to resonate within communities that for a generation had lost out on the increases in wages that had been seen elsewhere in the country.

Work by the Resolution Foundation suggested there is no relationship between recent changes in an area’s prosperity and how they voted at the referendum. While some areas that voted to leave the EU had seen a big increase in real hourly earnings, such as Christchurch in Dorset, others that voted to remain in the EU had recently experienced a sharp drop in hourly earnings, such as Rushcliffe in Nottinghamshire. However, further exploration at the aggregate-level suggested it was actually long-term entrenchment rather than recent change in the levels of incomes that tended to explain why support for Brexit was higher in some areas. Overall, it was areas where people tended to earn less that voted for Brexit even if these were not always the communities that had been the most badly affected in recent years. The implication is that ‘it’s the shape of our long lasting and deeply entrenched national geographical inequality that drove differences in voting patterns’.

The relationship between migration and level of support for Brexit

Another area of interest is the relationship between the vote for Brexit and migration, though current findings are mixed. Italo Colantone and Piero Stanig claim there is no evidence of correlation between support for Brexit and the proportion of immigrants or new immigrants. If anything, they argue, areas with more arrivals were more likely to vote Remain and areas with fewer arrivals were more likely to vote leave. But their claims are contested. Our earlier work did not find a positive relationship between support for leave and the ‘static’ level of immigration but we did find a positive relationship between the change in immigration and support for Brexit. After controlling for factors such as education, age and the overall level of immigration, communities that over the past decade had experienced an increase in migration from EU member states were somewhat more likely to vote for Brexit. For example, in Peterborough the estimated size of the EU migrant population increased by about 7 percentage points and 61% voted leave. Even though areas with relatively high levels of EU migration tended to be more pro-remain, areas that had experienced a sudden influx of EU migrants over the last 10 years were often more pro-leave. This finding is consistent with the argument that when it comes to the effect of immigration on the referendum what appears to matter the most is the experience of sudden population change rather than the overall level. Indeed, as Geoffrey Evans and Jon Mellon show, public concern about immigration as a political issue over time in Britain strongly tracks actual levels of immigration.

In summary, the findings of existing research are somewhat mixed and reveal a clear need to drill down to examine both the area and individual level, to which we now turn.

New research on Brexit examining dynamics of the vote

Support for Brexit by demographic group – personal finance, education, and attitudes and values

During the referendum and its aftermath a large number of polls were conducted which looked at public support for Brexit. Although many polls differed in terms of their estimated share of the vote for leave and remain they did tell a consistent story about which groups had voted leave. As with the aggregate analysis they found clear divides on age, education and ethnicity. Put simply, older, white and more economically insecure people with low levels of educational attainment were consistently more likely to vote for Brexit than younger people, degree-holders, minorities and the more secure middle- and upper-classes.

We can build on this work by exploring new data from the British Election Study (BES) Internet campaign study based on a very large sample of more than 31,000 respondents. This element was carried out before the referendum, during May and June 2016, and measures people’s intended vote rather than how they actually voted. While this is an online survey that is not as methodologically rigorous as face-to-face random probability surveys the overall results were reasonably close to the final outcome in terms of the result and variation across counting areas. The BES is also very helpful because the questionnaire includes a wide range of topics, including attitudes toward the EU, the referendum campaign, immigration, social and political values more generally and people’s backgrounds.

Figure 1 shows how support for leave varied among different demographic groups as the referendum neared.

There is a strong relationship between household income and support for Leave. People living in the poorest households were much more likely to support leaving the EU than those in the wealthiest households. In households with incomes of less than £20,000 per year the average support for leave was 58% but in households with incomes over £60,000 per year support for leaving the EU was only 35%. Those out of work were also far more likely to support Brexit than those in full employment – support for leave among the former was 59% but only 45% among the latter. Similarly, people engaged in low-skilled and more manual occupations were much more likely to support leaving the EU than those who work in and more secure professional occupations – on average the leave vote among the former was 71% but among the latter was only 41%.

Support for Brexit and personal finance

It is also worth exploring the relationship between how people feel about their own financial situation and support for leaving or staying in the EU. There is some evidence in the BES that people who thought that their household’s financial situation had got a lot worse over the previous 12 months were somewhat more likely to support Brexit than those who thought that the financial position of their household had improved. Support for leaving the EU averaged 58% among those who felt that their financial position had weakened compared with 44% among those who felt that their position had improved.

Support for Leave was also higher – by around 18 percentage points – among people from white British backgrounds than it was among people from Black and minority ethnic backgrounds (BAME) or other white backgrounds. Whereas 52 percent of people from a white British background supported Leave, just 34 percent of people from a BAME background and 31 percent of people from a white Other background (mainly European) did so. However, there was not much difference by gender. There were also clear age differences, with support for Leave among people aged over 65 years some 31 percentage points greater than support among people aged 18-24 years old. Although these age differences received a great deal of attention during the campaign - and are clearly important - (we examine the sources of these age divides later in the report) it is worth noting that age was not the only, or indeed the most important social divide.

Lastly, and consistent with other research, people without any educational qualifications were far more likely to support leaving the EU than those with postgraduate qualifications. Support for leaving the EU was a striking 75% among those who lacked qualifications but just 27% among those who had achieved the highest level of education – a difference of nearly 50 percentage points. It is this educational divide that is absolutely central to making sense of why the country voted to leave the EU, a point that we will return to.

We can get a clearer sense of which groups of people were most likely to support leave by examining the impact of these different individual characteristics simultaneously using a statistical technique called ‘logistic regression’. This technique allows us to examine the ‘independent’ impact of each variable on support for leaving the EU while controlling for each of the other variables under consideration. For example, we know that people who went to university tend to end up with better paid jobs than people who left school at 16. So in Figure 1 part of what we observe as an effect of income may in fact be due to someone’s level of education. To get round this we can examine both variables (and others) simultaneously. By examining education and income together, we can tell whether people with similar education levels but different levels of income differ in terms of their support for leave.

Our results are presented in Table 1 below in the data behind the analysis section. When we consider the role of age, sex, ethnicity and income together they suggest that people on low incomes were significantly more likely to support leaving the EU. There are also clear differences by ethnic background: people from BAME and – in particular - white other backgrounds were much less likely to support Leave than people from white British backgrounds. We also examined the impact of whether people were foreign born or not, but this did not have a significant independent association with support for leave, once we had controlled for ethnic background. Older people were also much more likely to support Leave than younger people. Furthermore, the groups in society that tend to be at higher risk of poverty, such as people who left school with a very low level of educational attainment, were much more likely to support Brexit than the well-educated.

Support for Brexit and education

When studying this data it is the effect of education that is particularly pronounced. We can illustrate this by calculating the predicted probability of voting to leave the EU for different groups, and while holding everything else constant. Support for leave was about 30 percentage points higher among those with GCSE qualifications or below than it was for people with a university degree. In contrast, support for leave was just 10 percentage points higher among those on less than £20,000 per year than it was for those on more than £60,000, and 20 percentage points higher among those aged 65 than it was for those aged 25.

The finding that education is so important should not come as a surprise. Past research on Euroscepticism in other countries by Armen Hakhverdian et al. has consistently identified education as a central and increasingly important factor to understanding public hostility to the EU. In a number of European countries education has also emerged as a new cleavage between the so-called ‘winners and losers’ of globalisation, according to Stubager. Hanspeter Kriesi et al. noted that the so-called ‘winners of globalisation’ are highly-educated and qualified people who possess the ‘convertible resources’ that are necessary to compete in a globalised society and take advantage of the opportunities that it presents. By contrast, the ‘losers of globalisation’, or the ‘left behind’, possess fewer skills and see their jobs outsourced or are challenged by the increased competition that has followed the inflow of EU nationals, noted Robert Ford and Matthew J Goodwin. However, education is often thought to matter in a slightly different way as well, and can act as a socialising agent that inculcates people with a more outward looking and liberal perspective on life, according to Hainmueller and Hiscox. It is this latter point that brings us to explore the role of attitudes and values in the vote for Brexit.

Support for Brexit and the role of attitudes and values

The role of values occupied a key position in the referendum campaign, from debates about sovereignty and national identity to the issue of attitudes towards immigration. Unsurprisingly, attitudes on these issues are closely related to the leave vote. Nearly 90% of people who thought immigration was bad for the economy supported leave, compared with just under 10% for those who thought immigration was good for the economy. Similarly, whereas 88% of people who thought that the country should allow fewer immigrants in supported Brexit the equivalent figure among people who wanted to keep immigration as it is was just 21%. People who feel ‘very strongly’ English were much more likely to say they would vote leave than anybody else (71 versus 36%) – and it was this narrow conception of national identity – rather than a broader sense of feeling ‘very strongly’ British that mattered most. There is also some evidence that those people who felt disillusioned with politics – and agreed with the statement that ‘politicians don’t care what people like me think’ were more likely to support leave than people who disagreed with the statement (70 versus 30%).

However, and as Eric Kaufmann points out, there are sharp divides in the level of support for Brexit between people who hold socially liberal views and those who hold socially conservative views. As Figure 2 shows, people in favour of the death penalty and harsher prison sentences in general, and who are against equal opportunities for women and homosexuals are much more likely to support leave – to the tune of around 50 percentage points. This suggests that an underlying differences in the values that people hold are important to making sense of why some people were attracted to vote leave.

It is important to make the point that these attitudes and values should not be seen in isolation. Rather, they relate strongly to age and education. Older people, those with few qualifications, who live on low incomes and work in manual jobs tend to hold more socially conservative attitudes. In contrast, those who are younger, more highly educated and work in secure jobs tend to hold a more liberal outlook. Attitudes also vary across different parts of the country, and people living in low-skilled areas tend to be more socially conservative, have a stronger English identity and feel more politically disillusioned than similar types of people living in high-skilled areas. Accounts of Brexit which attempt to frame social and attitudinal factors in opposition to each other are misplaced. It is not attitudes or demographics: it is both. In the next section we examine the extent to which these attitudes are able to account for the demographic differences that we observed. 

Moving on: Exploring the relationship between education, individuals and place

The findings so far suggest that education plays a very important role at community and individual levels. Education is important for providing individuals and communities with the sort of skilled workforce that is able to prosper and take advantage of the new opportunities that globalisation has to offer. Areas where there are a large number of people who are highly educated can be regarded as high-skilled communities. Clearly, there are sharp differences in the overall life and work experiences of people who live in areas that strongly supported Brexit compared with those that strongly supported remaining in the EU. Whereas over 70% of people in low-skilled communities like Tendring (which covers Clacton) voted for Brexit, over 70% of people in very highly-skilled communities like Cambridge voted to remain in the EU. There are also sharp differences between which groups of people voted to leave or remain. Whereas over 70% of people with no qualification voted for Brexit, over 70% of people with a postgraduate degree voted to remain. These differences point to deep divides in Britain, both geographically and socially. But how do these divides overlap? And what underpins them?

Were low-skilled people in high-skilled areas more likely to vote Leave because they felt unable to compete? Or were high-skilled people in low-skilled areas more likely to vote leave because they lacked the same opportunities to get ahead that meet high-skilled people in high-skilled areas? Or are the different geographical patterns we observe just a reflection of the uneven distribution of skilled and unskilled workers across the country?

We can begin to answer these questions by simultaneously exploring the influence of people’s backgrounds, such as their level of education, as well as the characteristics of the area in which they live, such as whether they reside in a community that is, overall, highly educated. To do this we can combine census data on the characteristics of the parliamentary constituency in which people live with survey data on their individual attitudes. We examine area at the constituency level rather than the local authority level as this allows us to examine a slightly more local area. The results of the analysis are presented in Table 2 below in the data behind the analysis section, but we will summarise the main findings here.

At the individual level we have already seen how white people, older generations and those with low educational attainment were more likely to support Brexit. However, our results also reveal how the type of place where these people live also matters. People who live in areas that are low-skilled, where the average level of education is low, were more likely to lend their support to Brexit than people who live in areas that are high-skilled, where the average level of education is high.

In more technical terms we also find a significant ‘interaction effect’ between a person’s level of education and the educational profile of the area where they live. The level of support for Leave among graduates varied much more than among those with low levels of education across different types of areas and different parts of the country. For example, whereas the level of support for Brexit among people with GCSE or below qualifications was 16 percentage points lower in high-skilled areas than low-skilled areas, it was over 30 percentage points lower among people with A Levels or a University degree. This points to a very important finding, shown in Table 3 in the data behind the analysis section below.

Graduates who live in low-skilled communities were more likely to vote for Brexit, and more similar to those with low education, than graduates who live in high-skilled communities (and who were, in contrast, very different to those with low education). For the purposes of illustration low-skilled areas are defined as places where 10% of the local population are university educated and high-skilled areas are defined as places where 60% of the local population are university educated. In the most low-skilled areas the difference in support for Leave between the low and high educated is around 20 percentage points; whereas in high- skilled areas the difference is just under 40 points. Crucially, this reveals how a geographic divide overlays the social divide that we outlined above. In communities that are low-skilled support for leave was much more evenly distributed across different segments of society than in communities that are high-skilled and where people are notably more polarised along education lines.

How can we explain this pattern? There are several plausible interpretations. One is to do with the role of place and the availability of local resources and opportunities. Even if people possess educational qualifications and skills, if they are stuck in left behind areas that are experiencing decline then they are less likely to be presented with local opportunities to use these skills and get ahead in life. Such an environment can fuel feelings of exclusion or marginalisation. Therefore, the left behind in Britain face a ‘double whammy’. On one level the left behind may become marginalised in our society because of their lack of qualifications and skills, which puts them at a significant disadvantage in a modern and increasingly competitive economy. But on a second level they may also be further marginalised because they lack the opportunities to get ahead within their local communities. Amid the modern economy and also as part of an increasingly diverse society in which rapid social and demographic change is the new norm, this makes it extremely difficult for the left behind to adapt and prosper.

Between these two groups – the low and high educated - are those with an intermediate level of education. People with an ‘A-level’ or equivalent level of education occupy the middle ground between those who have just a GCSE-level education and those who have progressed to benefit fully from a university education. Interestingly, it is people with A-levels who seem to be especially sensitive to their surrounding environment. Whereas in low-skilled communities those with A-levels or equivalent are very similar to those with low education, in higher-skilled communities their propensity to support Brexit is closer to people who have a university degree. This suggests that people with A-levels are more sensitive to their environment than the two groups at the extremes. It appears that, depending on the local opportunities they face, those with A-levels either fall into the ‘left behind’ group and become more supportive of Brexit or are able to get ahead in life and thus, like the high-skilled, become less likely to support Brexit. If there are not opportunities, then those with A-levels fall into the left behind with the low-skilled; but if there are opportunities then they can make progress alongside the more high-skilled.

To what extent do social attitudes account for the social and geographic divides that we have found? The results from our analysis are presented in Table 4, below in the data behind the analysis section. Support for leave was higher among those who support the death penalty; who feel a very strong sense of English identity; are politically disillusioned (measured by agreement with the statement: ‘Politicians don't care what people like me think’); and who favours reducing immigration. Controlling for these attitudes also helps to explain - at least in part - the social and geographic divides that we have found. In formal terms we can say that the impact of education, age and regional differences are mediated by the attitudes and values of people who live there since, when we control for attitudes, the magnitude of the effect for social and geographical factors is greatly reduced. Part of the reason why education and age matters is because of the distinctive world views of people with high and low education; and old and young people. People with high levels of education and young people tend to be more socially liberal and more open to immigration than people with lower levels of education and older people. Similarly, part of the reason why there are such sharp differences in support for leave across different areas of the country are to do with the distinctive values of people who live in low- and high-skilled areas.

Conclusions

In the aftermath of the vote for Brexit our research has revealed a country that is deeply divided along not only social but also geographical lines. There are three core findings.

First, income and poverty do matter. Groups of voters who have been pushed to the margins of our society, live on low incomes and lack the skills that are required to adapt and prosper amid a post-industrial and global economy, were more likely than others to endorse Brexit. Looking ahead it is likely that persistent and growing inequalities will strengthen this divide and that the country will remain divided on these issues for some time to come.

Second, while part of this story is about income and poverty our results reveal how educational divides matter more. After controlling for other factors support for leaving the EU was consistently higher, and significantly so, among those people with only a GCSE-level of education, or below. These differences by educational attainment were far more striking than differences by income level. Responding to this educational inequality should be a pressing priority for any government as it is restricting the opportunities that are available to some.

Third, where people live also played a significant role. The left behind groups, those who were the most likely to support Brexit, face a ‘double whammy’. While they are being marginalised because of their lack of skills and educational qualifications this disadvantage is then being entrenched by a lack of opportunities within their local areas to get ahead and overcome their own disadvantage. Unless this double whammy is resolved it will become increasingly difficult, if not impossible, for the left behind to keep pace with those voters who both have skills and are benefitting from the opportunities that high skill areas offer.

The result of the referendum, therefore, has thrown new light on deeper social, geographic and cultural divides that often lay hidden below the surface of our national conversation. Looking ahead, it seems likely that these stubbornly persistent and growing inequalities will strengthen. Both regional and individual disparities have pushed to the margins overlapping groups of voters, who live either in areas of decline or who live on low incomes and lack the skills that are required to adapt and prosper amid an economy that is increasingly built for those with skills, qualifications and resources. The more disadvantaged voters that turned out for Brexit are also united by values that encourage support for more socially conservative, authoritarian and nativist responses. On the whole, Leave voters have far more in common with each other than they have things that divide them. Over three-quarters of Leave voters feel disillusioned with politicians; two-thirds support the death penalty; and well over half feel very strongly English. Over one third of Leave supporters hold all three of these attitudes, compared to just 6 percent who do not hold any of them. This more liberal group of Brexit voters, therefore, constituted a very small part of the coalition for leaving the EU.

About this analysis

The research in this report was conducted by Matthew Goodwin, Professor of Political Science at the University of Kent, and Oliver Heath, Reader in Politics at Royal Holloway, University of London. Data from the British Election Study is drawn from the Internet Panel Wave 8. Further information can be obtained via the authors.

The data behind the analysis

Table 1: Support for leave: Demographic factors, logistic regression
Individual characteristics Coefficient Standard error
Age 0.02*** 0.00
Female -0.03 0.03
Ethnic Minority -0.26** 0.08
White Other -0.56*** 0.08
Education    
A Level -0.56*** 0.04
University Education -1.24*** 0.04
Household Income    
£20-39,000 -0.12** 0.04
£40-59,000 -0.17*** 0.05
£60,000 plus -0.39*** 0.06
Constant -0.26 0.08
N 19,903  

Notes: *** denotes p <0.001; ** p <0.005; * p <0.05. Reference categories are GCSE level education or below; Less than £20,000; white British

The coefficient refers to the log odds ratio. Values marked with an asterisk indicate that there is a statistically ‘significant’ difference in the likelihood of supporting leave between the group in question and the reference category, controlling for the other variables in the model. Values greater than zero indicate that the group in question is more likely to support leave than the reference group, holding all other factors constant; and values less than zero indicate that the group in question is less likely to support leave than the reference group, holding all other factors constant. 

Table 2: Support for Leave: People and places, logistic regression
Individual characteristics Coefficient Standard error
Age 0.02*** 0.00
Female -0.04 0.03
Ethnic Minority -0.27** 0.09
White Other -0.50*** 0.09
Education    
A Level -0.13 0.15
University Education -0.77*** 0.12
Household Income    
£20-39,000 -0.11* 0.04
£40-59,000 -0.14* 0.06
£60,000 plus -0.31*** 0.06
Area characteristics    
% Degree -0.01*** 0.00
% Age 65+ 0.03*** 0.01
% Born in UK -0.01*** 0.00
Cross-level interactions    
% Degree*A Level -0.02** 0.01
% Degree*Degree -0.02*** 0.00
Constant 0.91 0.23
N 19,903  

Notes: *** denotes p <0.001; ** p <0.005; * p <0.05. Reference categories are GCSE level education or below; Less than £20,000; white British

Table 3: Predicted support for Leave by education and area
education Area
Low-skilled (%) High-skilled (%) Difference
GCSE 70 54 16 points
A Level 64 29 35 points
Degree 49 18 31 points
Table 4: Support for Leave: People, places and values, logistic regression
Individual characteristics Coefficient Standard error
Age 0.01*** 0.00
Female 0.02 0.04
Ethnic minority -0.08 0.11
White Other -0.22* 0.11
Education    
A Level -0.37 0.19
University education -0.01 0.16
Household Income    
£20-39,000 -0.14* 0.05
  £40-59,000 -0.17* 0.07
£60,000 plus -0.20* 0.08
Attitudes    
Support death penalty 0.71*** 0.04
Strong English identity 0.74*** 0.04
Political disillusion 0.57*** 0.04
Against immigration 2.04*** 0.04
Area characteristics    
% Degree 0.00 0.00
% Age 65+ 0.02* 0.01
% Born in UK -0.01** 0.00
Cross-level interactions    
% Degree*A Level -0.02** 0.01
% Degree*Degree -0.02** 0.01
Constant -1.56 0.30
N 16,574  

Notes: *** denotes p <0.001; ** p <0.005; * p <0.05. Reference categories are GCSE level education or below; Less than £20,000; white British

Reference notes

Matthew J Goodwin and Oliver Heath (2016). ‘The 2016 referendum, Brexit and the Left Behind: An aggregate-level analysis of the result’, Political Quarterly, forthcoming

Technical notes

The data from the British Election Study does not take into account household size so is not a direct measure of failure to achieve a minimum living standard.

We have to be cautious when interpreting the data on occupation as the BES was unable to collect detailed occupation classifications for a large part of the sample. However, the broad trend is consistent with the data presented by Lord Ashcroft, who finds clear (though slightly less pronounced) differences of around 20 percentage points in levels of support for leave between the AB social grade (middle class) and DE social grade (working class). However, due to the large number of missing values on occupation we do not consider this variable in our multivariate analysis.