Abstract
The Leave majority recorded in England was decisive in determining the UK-wide referendum result. Brexit was made in England. We take this as a prompt to challenge the conventional Anglo-British mindset that animates most studies of ‘British politics’ and has shaped public attitudes research on the United Kingdom. We explore the persistence of distinctive Eurosceptic views in England and their relationship to English national identity prior to the referendum. We then model referendum vote choice using data from the Future of England Survey. Our analysis shows that immigration concerns played a major role in the Brexit referendum, alongside a general willingness to take risks, right-wing views, older age, and English national identity. Therefore, Brexit was not just made in England, but Englishness was also a significant driver of the choice for Leave.
Introduction
In the 2016 Brexit referendum, the United Kingdom as a whole chose to Leave the European Union (EU) by a narrow margin: 51.9%–48.1% of those who voted. There was, however, a much more decisive 7-point margin in favour of Leave (53.4%–46.6%) in England. Given that it is home to 84% of the United Kingdom’s population, the vote to leave in England outweighed substantial Remain majorities in Scotland (62.0%–38.0%) and Northern Ireland (NI) (55.5%–44.6%). While Wales also had a Leave majority (52.5%–47.5%), with less than 5% of the United Kingdom’s population, it did not play a decisive part in the overall outcome (see Table 1). Brexit was made in England.
Average support for remain in opinion polls, by territory, January–May 2016.
Source: Jeffery, 2016.
So we take England as our unit of analysis, asking why the commitment to leave the EU was so strong there. As will become clear, our answer (building on previous contributions by Henderson et al., 2016; Jeffery et al., 2014; Wyn Jones et al., 2012, 2013) has to do with the importance of national identities in England—Englishness and Britishness—in shaping attitudes to the EU.
These analytical choices—of England as a territorial unit of analysis and national identity as a key explanatory variable—might seem obvious, even prosaic. After all, academic research has long emphasised nationalism as a core component of Euroscepticism (Hooghe et al., 2002). Yet, to focus explicitly on England and its national identities goes against the grain of conventional understandings of UK politics. While the politics of NI, Scotland and Wales have long merited sustained attention from specialists (recent examples include Gormley-Heenan and Aughey, 2017; Hunt and Minto, 2017; McHarg and Mitchell, 2017), with few notable exceptions (e.g. Aughey, 2007; Kenny, 2014; Wellings, 2012), England does not receive the same treatment. Rather, the complexities of a state with four component units become simplified into the study of a ‘British politics’, a process which has the effect of veiling the characteristics and impact of England, the biggest part of the United Kingdom, at the same time as marginalising engagement with the other parts of the UK. This effect occurs because the dominant approach is based on what might be termed a triple effacement. First, NI is bracketed out as too difficult and different. British politics is then studied using data that are dominated by respondents based in England, hiving off Wales and Scotland as the domain of country specialists. Third, data that are in fact overwhelmingly English are not analysed on an all-England basis—indeed, the country’s name is seldom even mentioned. The effect is to efface England too through the dominant British politics approach to the analysis of UK politics. Through this triple effacement, forgetting NI, pushing Scotland and Wales to the analytical margins, and veiling England, we end up analysing the United Kingdom as a fictive country: Anglo-Britain.
While academic analysis has been slow to recognise the significance of politically mobilised national identities in England for the United Kingdom as a whole, the antennae of some political leaders have been more sensitive. David Cameron’s 2015 general election campaign was nuanced to appeal to national sentiment in England (Jeffery et al., 2016), as was his doorstep speech the day after the 2014 Scottish independence referendum. Even earlier, the then leader of the UK Independence Party (UKIP), Nigel Farage, had recognised the potential for mobilisation in and for England, as Ford and Goodwin (2014: 90–91) have reported: There is a rebirth of identity politics in this country … [w]e’ve seen it in Scotland and it’s happening in England but no one has noticed. It’s little things. It’s the turn-out at Remembrance Day parades. They go up every year! A younger generation, an under-45 generation is hungry to know about their history and what their grandparents did and where they come from. It’s very interesting. It’s actually bloody happening.
On 23 June 2016, to paraphrase Farage, it actually bloody happened: English national identity was mobilised by the Brexit question. This article examines how. In a first step, we review recent work on the Brexit referendum in which English and other national identity variables have featured (e.g. Clarke et al., 2017a; Clarke et al., 2017b; Goodwin and Heath, 2016; Goodwin and Milazzo, 2017; McAndrew, 2016; Ormston, 2015). It finds that they reveal some of the problems of the ‘mainstream’ British politics approach treatment of England. Most appear to be reluctant to treat England as a unit of analysis and are uncertain about how to explore the role of national identities in England. Second, we explore the pattern of opinion in England on European integration, comparing it with attitudes in other parts of the United Kingdom. We show that Eurosceptic attitudes have a longer history of distinctive links to England and Englishness; these connections did not just appear out of the blue in June 2016. In the final section, we present a model of the Brexit referendum in England which shows how Englishness in England, as distinct from Britishness, combined with a number of other key variables to explain why England voted to Leave, taking the wider United Kingdom with it.
Analysing England, national identities, and Brexit: What we know so far
Given its historic importance, it is not surprising that the Brexit referendum is generating significant academic interest. A number of studies have already appeared which touch on the main concern here, namely, the centrality of England and English national identity to understanding the overall result. That work reveals two problems: how to treat England as a unit of analysis, and how to conceptualise national identities in England.
England as a unit of analysis
Most of the relevant studies (McAndrew (2016) and Ormston (2015) are exceptions) take Great Britain as their unit of analysis. This choice immediately excludes one part of the United Kingdom—NI. This exclusion is unfortunate given NI’s Remain vote and its prominence in the Brexit negotiations. Adopting a Britain-wide focus can also disguise the extent of territorial differences within Britain. This occurs, for example, in analyses where the component parts are compared to Britain-wide reference values. Distinctive attitudes to the EU (and other issues) in Scotland have been assessed by comparison with Britain-wide data (see, for example, Curtice, 2013), but since Scottish respondents are also part of the British comparator sample, this approach understates Scottish distinctiveness. By the same token, unless the English sub-sample is disaggregated, the approach can also veil England’s distinctiveness within the British or UK-wide figures, leaving England ‘hidden from view’ (Jeffery and Wincott, 2010; Wimmer and Glick Schiller, 2002).
Britain can be a valid unit of analysis, if it is carefully considered and specified on a basis that reflects diversity across the island. But too often, a ‘whole of Britain’ (or ‘Anglo-Britain’) approach is adopted by default rather than as a considered choice—dominated by English data, without ever analysing England as such. There is a strong case for approaching British politics through ‘bottom-up’ analysis of its three component nations rather than (or as well) as a single unit of analysis (and indeed to add NI and consider the United Kingdom as a whole). One advantage of such an approach would be to explore Britishness in a differentiated way; British national identity may not be uniform in all parts of the United Kingdom (as Bechhofer and McCrone, 2010 recognise). Of the studies explored below, only Ormston adopts the bottom-up approach.
To take a bottom-up approach requires careful attention to the data we collect and use as political scientists. A British survey frame may not ensure sub-samples sufficient for rigorous analysis in Scotland or Wales (and excludes NI altogether). Fortunately, both academic and commercial research increasingly produces data on Scotland, Wales and NI as discrete units of analysis. But it has also long been possible to extract robust sub-samples for England from Britain-wide surveys, given its population share. This possibility exists for all the main survey vehicles in regular use over recent decades by political scientists. That few have taken England as a unit of analysis is therefore a matter of choice rather than a necessity driven by the availability of data.
England and its national identities
Several Brexit referendum studies do use an English identity variable, although national identity is typically not their core concern. Rather, they set out causal hypotheses focused on such other issues as Anglicanism (McAndrew, 2016), poverty and place (Goodwin and Heath, 2016), or attitudes towards immigration (Goodwin and Milazzo, 2017). In general, these analyses are conducted in standard multivariate regression model form. In addition to the variables of particular interest, the models usually test a combination of socio-demographic variables, party identification measures, campaign cues, as well as a range of other values and attitudes, including national identity. The studies all use British Election Study (BES) data and, with the exception of McAndrew, take ‘Britain’ as their unit of analysis. The Clarke et al. (2017a, also 2017b) book-length study presents a more theoretically ambitious analysis, rooted in valence politics. Again, Britain is the unit of analysis, accessed mainly through data from the Essex Continuous Monitoring Survey (ECMS). Ormston (2015) is the only study that differentiates across the United Kingdom’s nations and national identities, using data from a variety of sources, although her analysis is not developed in a multivariate model.
National identity variables are used in different ways in these studies. McAndrew (2016) deploys the ‘Moreno scale’ (used to explore the relationship between two national identities that coexist and can overlap in the same territory) to identify those giving priority to English over British identity. She finds the ‘English’ are more likely to support Brexit. Goodwin and Heath (2016) focus on identity intensity, using a measure of ‘strong English identity’ as one of four ‘attitudinal’ variables (alongside support for the death penalty, political disillusionment, and opposition to immigration) in a model that uses socio-demographic, education, income, and area characteristics. All four attitudinal variables significantly predispose respondents to support Brexit and weaken the significance of other variables. Immigration has the largest impact, with strong English identity second. Immigration is the primary concern for Goodwin and Milazzo (2017). They look both at the choice for Brexit and at switchers from ‘Remain’ to ‘Leave’ between 2014 and the referendum. They use intensity variables for English, British and European identity. For the Brexit choice model, they find that both strong English and strong British identities are linked to voting Leave and that European identification has a larger pro-Remain effect. In their ‘switchers’ model, which finds significant impacts from attitudes to immigration, Englishness is not significant, while strong British identity is linked to switching to Leave. European identification has a negative impact on making this switch (Goodwin and Milazzo, 2017). In none of these studies, it should be noted, is there much in the way of discussion or analysis of these identities and why they might be significant. They tend to be variables that ‘pop up’ in the analysis, rather than driving it.
Clarke et al. (2017a, 2017b) do go further. They delve into the literature on public attitudes on European integration to analyse the relative effects of ‘calculation, community and cues’ (Hooghe and Marks, 2005) in explaining Brexit choice: calculations of cost and benefit, feelings of national identity and cues given by politicians. Applying a valence politics approach, they focus on the economy and immigration in accounting for Brexit. In exploring the impact of ‘community’, their territorial unit is Britain-wide. Using an open-choice national identity question (allowing respondents to claim one or more national identities simultaneously), they examine the effects not only of English but also of Welsh and Scottish identity. British identity is used as a reference category, not included directly in the models. Except for Englishness, all national identity variables were associated with Remain, but only Scottish identification proved to be significant. However, English identifiers were more likely—at a statistically significant level—to emphasise the benefits rather than the costs of leaving the EU (Clarke et al., 2017a: 161–168).
The focus on identities outside England is also a feature of Ormston’s (2015) pre-referendum analysis. She explores the effects of British as well as English, Scottish, and Welsh identities on Brexit voting intention in different parts of Great Britain. Uniquely among these studies, Ormston gives (passing) attention to opinion polls in NI. Her analysis is specifically focused on nations and national identities, drawing mostly on data from BES, British Social Attitudes (BSA), and its stablemate Scottish Social Attitudes (SSA) survey. It reveals significant differences in attitudes towards Brexit by nation. The impact of national identity also varies: English, Welsh, and Scottish identities have different relationships to Brexit, and the impact of British identity is not uniform across Britain, a feature of British identity we have noted in previous publications (Jeffery et al., 2014).
Two issues arise from this brief survey. The first is that it cannot simply be assumed that British identity has a uniform meaning across all the territories of Britain (let alone the United Kingdom). Yet (apart from Ormston) only McAndrew recognises this point by restricting her analysis of ‘Moreno’ identities to England. Elsewhere, undifferentiated British identity is used across Britain: to compare ‘British’ effects with those of English identity (Goodwin and Heath, 2016; Goodwin and Milazzo, 2017) or as a reference value against which to measure variation by English, Welsh, or Scottish identity (Clarke et al., 2017a). That Britishness may mean different things, and have different effects, in different parts of the United Kingdom is not recognised.
Second, assumptions about how national identity relates to communities at different geographical scales are problematic. For example, Goodwin and Heath (2016: 12) describe people who feel very strongly English as having a ‘narrow conception of national identity’, without further elaboration as to what ‘narrowness’ might mean. Conversely, feeling very strongly British is described as ‘broader’. Clarke et al. (2017a, 2017b) echo what we might term a ‘width-based’ conception of identity, distinguishing between ‘exclusive’ and ‘inclusive’ identities. Compared ‘with those identifying themselves as British’, they argue that: we expect those thinking of themselves as English or Welsh will be less favourable towards EU membership, whereas those who see themselves as European … will be more favourable. This is because the former identities are narrower than a more inclusive identity of being ‘British’. (Clarke et al., 2017a: 160; emphasis added)
Tellingly, however, Clarke et al. do not adhere consistently to this ‘width-based’ conception. Indeed, they suggest the opposite immediately after its introduction, claiming that the ‘recent upsurge of nationalism’ in Scotland ‘flips’ the relationship between territorial scale and narrowness/exclusivity ‘since many Scottish identifiers see EU membership as an attractive alternative to remaining in the UK’ (Clarke et al., 2017a: 160). Elsewhere, noting Haesly (2001), they suggest that ‘evidence for … identity influences … is mixed, since Scottish and Welsh identities have been shown to enhance support for European integration’ (Clarke et al., 2017a, 64). So ‘narrowness’ and a smaller territorial scale matter, except when they do not. There is a conceptual mutability here which suggests that thinking about the impact of national identities on Brexit—in England or elsewhere—is still at an early stage. Yet, in the English case, there is, as the next section shows, plenty of evidence that national difference, related to national identity, is not a new phenomenon.
The EU and England, 1993–2017
As has already been noted, even if Britain-wide surveys are not designed to focus on England as a unit of analysis, extracting a viable sample of respondents domiciled in England is perfectly possible. And for those who have looked, England’s distinctive attitudes on the EU could be found long before June 2016 (Henderson et al., 2016). In 1975, of course, support for the United Kingdom’s continued membership was strongest in England and then Wales and weakest in Scotland and NI. Since then, these rank orders have changed. Figure 1 compares BSA time series data from the early 1990s to the early 2010s for England and Scotland, supplemented from SSA at some time points. It shows the proportion of those who wanted to leave the EU or to significantly reduce its powers. Attitudes in England and Scotland move in broadly similar directions and at times come close to convergence. The level of hostility to the EU in Scotland, however, never equalled or exceeded that in England. The persistence of this gap from before devolution suggests that there is more to this difference than a Scottish instrumentalisation of ‘Europe’ ‘as an attractive alternative to remaining in the UK’ animated recently by the prominent issue of Scottish independence (Clarke et al., 2017a: 160). In 1993, there was a 15-point gap between English and Scottish support for leaving the EU or reducing its powers. Over the next 20 years, the gap widened slightly, but devolution did not bring a sea change in attitudes to the EU: over a long period, there were clear differences in attitudes to the EU in England and Scotland.
Support for leaving the EU/reducing its powers in England and Scotland.
During the run-up to the Brexit referendum, data generated by commercial pollsters showed a similar pattern of territorial differentiation. Admittedly, pollsters did not emerge from the referendum with enhanced reputations: in its immediate run-up, only two of the eight major polling houses called the outcome for Leave. The average of these polls was a 52%–48% margin for Remain (although Internet polls generally seem to have been more accurate than phone polls—for a helpful discussion, see Clarke et al., 2016). These predicted results from commercial pollsters were based, however, on Britain-wide samples. From early 2016, pollsters began releasing results with reliable sample sizes for all four parts of the United Kingdom. Table 1 records the average Remain voting intention for each one, with 4 polls in NI, 6 in Wales, 20 in Scotland, and 16 in England. All polls in Wales and NI and almost all in Scotland were Internet-based. In England, there was a roughly even mix of Internet (nine) and telephone (seven) surveys; the different methods produced significantly substantially different results. Table 1 records both Internet and telephone poll averages for England. It weights the respective averages by relative population share to produce notional, bottom-up projections of a UK-wide vote.
These figures are only heuristic: they are based on a different number of polls in each place, conducted by different pollsters using distinct ‘house’ methodologies, and the population weighting calculations assume uniform cross-UK turnout. But the findings remain interesting. In Scotland and NI, Remain was clearly ahead. Wales was close to evenly split. In England, different polling methods produced a 5-point difference in projected outcome. Findings using phone polling in England projected a comfortable Remain majority UK-wide, but internet polling projected a knife-edge outcome. England would (just) have been held in the EU against its will on the calculations in Table 1, and very small opinion movements in England could have tipped the outcome the other way. In the event, all parts of the United Kingdom recorded lower support for Remain, than the pre-referendum polls suggested. England and Wales voted to Leave and Scotland and NI to Remain, leaving the divided outcome some had foreseen (Henderson et al., 2016: 190).
There are two crucial points to draw here. First, the pattern of opinion in England—with 84% population share—was decisive. Second, opinion in England has differed from that in Scotland for some time (we lack robust data to be sure about Wales and NI). Using available data to unveil England—in other words, ‘taking England seriously’—whether on the long view using BSA data, or in taking a bottom-up view (at least of internet polling data) in the more immediate run-up to the referendum, would have revealed this distinctiveness (as demonstrated in Jeffery et al., 2014).
Insights from the Future of England survey
Until recently, treating England as a discrete unit for public attitudes analysis has been rare. In the early 2000s, a flurry of England-only surveys was conducted (see Curtice, 2006). They fizzled out after 2004 when the policy of establishing elected regional assemblies in England ended. To fill this gap, the ‘Future of England Survey’ (FoES)—a series of studies of political dynamics in England (Jeffery et al., 2014; Wyn Jones et al., 2012, 2013)—was launched in 2011. The FoES built on earlier studies of sub-state attitudes and identities—Citizenship after the Nation-State (Henderson et al., 2013a, 2013b) and Towards a Regional Political Science (Jeffery and Schakel, 2013). It drew on an overarching critique of the ‘methodological nationalism’ which can hide sub-state politics from view (Jeffery and Wincott, 2010).
Distinctively, the FoES found a range of discontents associated with strong or exclusive English national identity: the perceived unfair advantages Scotland gets from devolution, the absence of an institutionalised political voice for England, immigration, and European integration. The FoES showed that English national identity and these accumulated discontents have been disproportionately linked with political parties on the right, both the Conservative Party and UKIP. By contrast, in England, respondents with a strong or exclusive British identity tended to more content with the state of the two political unions—United Kingdom and EU—of which England formed a part and tended also towards support for left of centre parties (on ‘the two unions’ see, in particular, Wyn Jones et al., 2013).
The FoES uses multiple measures of national identity. One of these, the Moreno bipolar identity scale, asks respondents to ‘weight’ their relative attachment to Englishness and Britishness across five categories. Table 2 uses 2012 FoES Moreno data to analyse two key EU-related questions: whether the United Kingdom’s membership was a ‘good’ or a ‘bad’ thing, and Remain/Leave options in a possible future referendum.
National identities and EU membership in England in 2012 (%).
Source: Future of England Survey 2012 Results taken from Wyn Jones et al. (2013: 19).
EU: European Union.
Neither/Don’t know/Wouldn’t vote responses not reported.
The results are strikingly linear: the more English respondents felt, the more they saw EU membership as a bad thing, and, if given the chance, they would vote Leave. Conversely, feeling mostly or exclusively British in England pushed respondents clearly (though not as strongly) towards support for EU membership. To test that these identity effects were distinctive to England, the 2014 FoES was extended to ask the relevant questions in Scotland and Wales too (see Table 3). Towards the British identity end of the Moreno scale, the balance of opinion on EU membership was fairly consistently positive across England, Wales and Scotland (with one exception—the ‘more British than Welsh’). English ‘Brits’ were the most positive. Towards the ‘sub-state’ (or, perhaps, ‘narrow’) end of the scale, we see striking differences. Here, the balance of respondents in Scotland was clearly pro-EU, evenly split in Wales, and in England decisively against membership.
National identities and EU membership in England, Wales, and Scotland in 2014 (%).
Source: Future of England Survey, 2014.
EU: European Union.
Neither/Don’t know/Wouldn’t vote responses not reported.
Moreover, the extent of variation across the Moreno scale differs sharply between the three nations. In Scotland, variation on the ‘good thing’ question across identity categories ranges from 41% to 45%, in Wales from 31% to 37%, but in England, it is a striking 15%–52%. The ‘Remain’ question shows a similar pattern, ranging across 47%–53% in Scotland, 34%–45% in Wales, but 20%–56% in England. National identities shape attitudes markedly and in a polarised manner in England, hardly at all in Scotland, and a little more in Wales.
How Brexit was made in England
These findings—on national identities and attitudes to the EU in England, and their distinctiveness to England—shaped our approach to the 2016 FoES, fielded in the immediate run-up to the Brexit referendum. Conducted by YouGov, the survey involved 5103 respondents between 10 and 21 June 2016. This real-time snapshot as voters were finalising their decision found 52.9% support for Leave and 47.1% for Remain (the actual outcome in England was 53.4%–46.6%). As in 2012 and 2014, the 2016 FoES data show a clear relationship between Moreno identity categories and voting intention (Table 4). Mostly or exclusively, English respondents were much more inclined to vote Leave than those prioritising a British identity.
National identities and 2016 referendum voting intention in England (%).
Source: Future of England Survey, 2014.
Results are column percentages. Results might not sum to 100 due to rounding. Neither/Don’t know/Wouldn’t vote responses are excluded.
To facilitate further statistical analysis, we have collapsed identity categories, allowing us to compare those who prioritise English identity (either English only or English more than British—column (c) in Table 4) with those who give equal weight to both identities (column (d)) and those who prioritise Britishness (column (e)). A clear majority of those who prioritised their English identity (73%) said they intended to vote Leave. For those who held both identities equally, there was a roughly even split on Brexit preferences, with 49% opting to vote Leave. Just over one-third (35%) of those who prioritised their British identity said they intended to vote this way. Brexit was ‘made in England’ and English identifiers played a key role in making it.
To test these national identity effects further, we have developed a model of referendum vote intention that includes identity variables alongside others, including those highlighted elsewhere in post-referendum analysis, in particular variables that allow us to explore whether or not it was ‘left-behind’ (Goodwin and Heath, 2016) voters—older, less well-educated, less affluent, and lower class—who opted to vote Leave. 1
Building on earlier FoES work, and some other post-Brexit analyses, we include variables for both English and British national identity, expecting the former to be associated with Leave and the latter with Remain. We construct two binary variables from the Moreno scale: individuals are coded 1 if they describe themselves as English not British or More English than British and 0 otherwise and structure a binary measure of British identity in England in the same way. These measures do not explore identity intensity. Three quarters of FoES 2016 respondents felt themselves to be strongly English (at points 7–10 on a 0–10 scale) and 76% strongly British. This high overlap among strong identifiers in each category makes isolating their individual effects difficult. The Moreno measure pushes respondents to balance the relative weight of these identities, and allows us to capture their effects.
We deploy other variables to delve more deeply into individual feelings about England and Englishness. Two ‘nostalgia’ variables draw on questions about the British past. Although we are sensitive to the danger of subsuming England into Britain, our analysis elsewhere (Henderson and Wyn Jones, 2017) suggests nostalgia for a Britain of yore is peculiarly English. The first nostalgia variable captures the belief that ‘Britain’s best time was in the past’. The second measures agreement with the statement ‘Young people today don’t have enough respect for traditional British values’. Other variables are designed to reveal a sense of national difference in England and measure agreement with the following statements: ‘English culture is not valued as highly in England as other cultures are’ and ‘In England our values make us different from the rest of the UK’. These variables help us to unravel how English identity, attitudes to the past or perceptions of a distinctive English national community may have influenced Brexit vote intention.
Other studies show a strong relationship between support for Brexit and attitudes to immigration. Indeed, control over immigration was at the heart of the ‘taking back control’ message projected by the Leave campaign. Like Goodwin and Milazzo (2017), we have included immigration variables to tap different dimensions of the issue. We include measures of support for reducing not only immigration numbers but also attitudes to cultural dilution or assimilation (measured as support for the view that immigrants dilute British culture or that immigrants should blend in). To these, we add measures of more instrumental concerns about the pressures immigration may place on public services or that it might act as a drain on the economy. These variables allow us to explore not just the general impact of the immigration issue but also whether cultural or material concerns about immigration weighed more heavily on voters’ minds.
We include a set of more general attitudinal variables. As the United Kingdom’s relationship to the EU has often been framed as about (warding off) interventionism and regulation, those who describe themselves as right wing, and are therefore less supportive of state intervention, may be more supportive of Leave. Other research has shown that attitudes to risk and efficacy can influence referendum vote choice. Individual tolerance for risk is associated with greater support for change regardless of the particular change on offer (see, for example, Nadeau et al., 1999). And, as Brexit has been framed as an opportunity to ‘take back control’, we might expect individuals with a low sense of political efficacy—that is those who feel powerless in the democratic process—to be more pro-Brexit.
The results are set out in Table 5. The reported statistics are odds ratios for logistic regression. Any number above 1 indicates an increased likelihood of voting Leave. We have run the model in blocks, with demographic variables, followed by identity variables, those tapping a sense of nostalgia and national distinctiveness, attitudes to immigration and our other attitudinal variables added to subsequent models.
Modelling referendum vote intention.
Source: FoES (2016), n = 2026.
Reported figures are odds ratios from logistic regression.
p < .1; **p < 0.05; ***p < 0.01.
The results confirm some of what we expected. Demographic predictors such as age, education and social class matter, and in the predicted direction. Older voters and those in social classes C2DE are more likely to vote Leave, while those with university degrees are more likely to say that they will vote Remain. When we include additional variables, however, only age and social class remain significant predictors of support for Brexit, and in the fully specified model (5), only age exerts a significant effect. This result gives some limited support for the ‘left-behinds’ analysis of Goodwin and Heath (2016).
Our identity variables suggest that English and British identity influence Brexit vote intention in different ways. English identifiers are significantly more likely to say they will vote Leave and the effect is unchanged in the fully specified model. Feeling British is initially a predictor of support for Remain, but the addition of subsequent variables dilutes this effect. In short, English identity and British identity do not exert the same level of influence on attitudes to EU membership, nor do they work in the same direction. These different effects chime with the others’ results on the impact of English identity (Goodwin and Milazzo, 2017; McAndrew, 2016), although they suggest more caution may be needed in the treatment of British identity.
Turning to the other measures of Englishness, each is initially a significant predictor of support for Leave, especially the view that English culture is not as valued as others. When we add additional variables, however, the effects dampen, disappear or even flip in the direction of their effect, although it is worth noting that the effect of national distinctiveness survives the addition of immigration variables. These results suggest that what people in England understand by their English identity is less fixed and inward-looking than the language of ‘narrowness’ and ‘exclusivity’ used by others might suggest.
Our variables on immigration confirm our expectations as well as the previous findings of other researchers (Goodwin and Milazzo, 2017): concern about immigration boosts support for Leave. However, we identify additional nuance. Support for reducing the numbers of migrants is a significant predictor of Leave vote intention, as is the concern that immigrants dilute British culture, although attitudes to assimilation are not significant. Attitudes on the effect of migration on public services and the economy exert a stronger role. Citizen concerns over immigration appear to be more strongly connected to Brexit through perceptions of material circumstances than cultural aversion to immigrants.
General political attitudes also influence vote choice. Those who describe themselves as right wing and those tolerant of risk are more likely to incline towards supporting Leave. So too are those lacking a sense of efficacy. Right-wing beliefs and attitudes to risk exert a greater effect than any single measure of attitudes to immigration. Right-wing effects may reflect the dominance of Conservative Party and UKIP figures in the Leave campaigns. The efficacy impact on Leave voting—which echoes Goodwin and Heath’s (2016) findings on ‘political disillusionment’—is ironic given the claims about Brexit returning powers to Parliament, given that the same Parliament is perceived ‘not to care’. The effects of risk echo results in other referendums, where general risk aversion has served to dampen support for change in a case where the future is largely unknown (Henderson et al., 2014; Nadeau et al., 1999). Here, those with a higher risk tolerance, even amid the unknowns of Brexit, are most disposed to Leave voting.
These results echo some findings of other scholars, notably about immigration, and to an extent also about the preferences of ‘left behinds’. They suggest some modification to the standard interpretation in that they show relatively weak effects from social structural variables and much stronger influences from attitudes and values. They also highlight the different dimensions through which immigration attitudes play a role.
The results of our analysis also underline that national identity is a key element of the Brexit puzzle. So far, other studies have left it at that: national identity is one of a mix of relevant variables identified, but not a major focus for analysis. Our research shows that the two dominant national identities in England are not synonyms and that they do not exert the same level or direction of influence on Brexit preferences. They also show that some of the assumed avenues by which English national identity might influence attitudes to the EU—nostalgia and a perception of England as a distinct national community—are largely absent.
Conclusion—Taking England seriously
Brexit was made in England because of England’s population weight in the United Kingdom. And England’s choice for Brexit was driven disproportionately by those prioritising English national identity. We know from our earlier work that English national identity is a cluster point for other attitudes and concerns. These include hostility to European integration, the sense of absence of political voice, concern about immigration, and support for parties of the right. The model in Table 5 finds echoes of those findings in the drivers of Leave voting it identifies in England. These results also echo the broad findings of other work on Brexit, especially on immigration and to a lesser extent on the ‘left behinds’.
In recent years, Englishness often has been linked to a ‘narrow’/’exclusive’ imagery around English national identity. This image has shaped academic thinking and has a popular echo in the stereotypical, culturally insular, even intolerant ‘white van man’ draped in a flag of St George. We find some evidence of concern about the undervaluing of English culture and, perhaps linked, desires for cutting immigration and worries about its impact on British culture. But the impact of many of these variables washes out once further attitudinal variables are added to the model. Moreover, if Leave campaigners blew their dog whistles around images of vast queues of swarthy immigrants, they look to have been blowing in vain: concern about immigration seems to have tapped perceptions of its material impact more powerfully than its cultural dimensions.
Leave supporters across identity groups are more likely to feel that their representative institutions do not care about them. Their choice to ‘take back control’ may well prove to be a double-edged sword for a UK Parliament that is not held in high regard. This issue may be especially challenging in and for England, that part of the United Kingdom which lacks the meaningful devolved institutions that provide an alternative and more accessible route for political engagement. So while the people of England have spoken about the EU, their focus may now train uncomfortably on the UK Parliament. As G. K. Chesterton put it, ‘But we are the people of England; and we have not spoken yet. Smile at us, pay us, pass us. But do not quite forget’.
National identity is a complex social phenomenon, particularly in plurinational states. Even when not strongly politicised, it often has a significant social quality. Political nationalism is a potentially powerful force, which sometimes flares up temporarily and in so doing can alter politics and reshape institutions. But the social realities of nationalisms do not only influence politics through short-term mobilisations—they are also among the most fundamental and profound factors structured into our political institutions, thereby shaping political life within, between, and beyond states.
These reflections suggest that the political significance of national identifications in England is an evolving phenomenon, one not closed off by the decision for Brexit. English nationalism, in particular, may play in other ways into the future of UK domestic politics. Political science in the United Kingdom will need to work hard to meet this challenge. It is no overstatement to say that the Brexit referendum outcome was a surprise to the great majority of political scientists working on UK politics. It was a surprise because of the entrenched assumptions that continue to be reproduced about how to study UK politics, including a routine failure to take England as a relevant unit of analysis. A ‘British Politics’ tradition has veiled trends distinctive to England by taking Britain unreflectingly as its unit of analysis. However, as we have shown, it has long been straightforward to unveil this England, understand its distinctive views on European integration, and on that basis to foresee that there was a strong probability of a Leave vote. Had there been a clear understanding of this English constellation of attitudes and values in the years preceding the referendum, the calculations made by politicians and political parties about the question of EU membership could well have been different.
So we conclude with a call, now, even if one horse has bolted, for UK political science to take England seriously (Jeffery et al., 2014). The early wave of post-referendum work discussed here makes a start. But it is only a start, with little clarity yet about how to look at England and how to distinguish it from the Britain of which it forms the major part. The empirical base for a research programme on England is easy to construct, much of it from data that is already available. What is missing is the conceptual work to capture the multi-level political dynamics of that part of the United Kingdom which has, as yet, lacked a formal institutional footprint.
Footnotes
Acknowledgements
The authors would like to thank Alena Drieschova, Jac Larner, and Ed Gareth Poole for comments on previous drafts.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.