The generational decay of Euroscepticism in the UK and the EU referendum

ABSTRACT

A prominent feature of media coverage during the UK’s referendum on European Union (EU) membership was the stark difference between the pro-EU young and their Eurosceptic elders, widely assumed to reflect a generational divide. The positive relationship between age and hostility towards the EU is well established in academic research, however only Down, and Wilson [(2013). “A rising generation of Europeans? Life-cycle and cohort effects on support for ‘Europe’.” European Journal of Political Research 52: 431–456] have considered whether this reflects a generational or life-cycle effect. While their research confirms that there is such a generational effect, their capacity to explain it is limited. This study utilizes data from Britain and builds on previous attempts to identify and explain generational trends in Euroscepticism, bridging it with studies on individual-level determinants of hostility towards the EU, providing the most detailed assessment of the extent and causes of generational differences in Euroscepticism to date. The results confirm that today’s young people are the most supportive generation of EU membership, caused by a combination of factors including their experience of the EU during their formative years, their relationships with domestic political institutions, and their access to education.

Introduction

The referendum on UK membership of the European Union (EU) produced one of the biggest shocks in modern British politics. A prominent theme of the campaign and fallout of the result has been that of conflict between the largely pro-EU young and their Eurosceptic elders (WISERD 2016).¹ This was widely assumed by the media and politicians to reflect a generational divide, i.e. something about the environment in which young and older generations were socialized made them more or less supportive of the EU. Previous research has extensively documented the positive relationship between age and Euroscepticism, but most studies do not employ methods capable of separating cohort and life-cycle effects, meaning they cannot reliably identify the nature of this relationship. Only Down and Wilson (2013, 2017) have addressed this issue, and showed that younger generations are typically more supportive of EU membership but become more Eurosceptic as they age.

Considerable questions remain, however, about the cause of the generational effect. Down and Wilson (2013) attribute it to younger generations being socialized during a time of greater EU integration, causing them to view a progressively stronger EU (and weaker national government) as normal. Their analysis did not consider in detail, however, a range of traits known to affect Euroscepticism identified in the extensive literature on its individual-level causes, largely because of limitations in their data. Consequently, their research potentially over-estimates the importance of EU integration as the source of generational differences in Euroscepticism.

This article addresses this challenge using age–period–cohort (APC) analysis to identify generational differences in Euroscepticism and explore the role of theories of individual-level Euroscepticism in explaining them. The study overcomes the limitations of multi-national data sets (such as Eurobarometer) by focusing on the single national context of Britain and making use of the extensive, detailed survey data on the attitudes of British citizens. The similarities in trends and causes of Euroscepticism between Britain and other member states mean that the conclusions are assumed to be generalizable, though clearly a direction for future research is proving that this is the case. Another advantage of focusing on Britain is that overlap between Euroscepticism and support for leaving the EU in the recent referendum means the conclusions of this research can aid with explaining the result of one of the most dramatic referenda in British political history.

This study confirms that there is a generational effect, with today’s young people at the leading edge of a generational decay in Euroscepticism. There is also a stronger, countervailing life-cycle effect, in which people become more Eurosceptic as they age. The declining attachments of younger generations to domestic institutions that either promote a Eurosceptic outlook (such as the Anglican church) or are focused on influencing and controlling the national government (such as the Labour or Conservative parties) is an important driver of the generational decay, as are rising levels of political sophistication and weakening concern with traditional conceptions of national identity. While the contribution of these processes to explaining the generational differences is limited, the paper nonetheless concludes that generational differences in Euroscepticism are not solely caused by differences in the level of EU integration during the formative years.

The paper begins by outlining the impressionable years theory of political socialization that explains how generational trends in political characteristics form, before reviewing theories of individual Euroscepticism. It then details the research design, focussing on APC analysis and discussing the political generations identifiable within the British electorate. The final sections present the results and conclude by considering the implications of these findings for future studies of Euroscepticism, as well as for explaining the result of the UK’s referendum on EU membership.

The impressionable years and causes of generational trends in Euroscepticism

Claims of generational differences in Euroscepticism are based on the impressionable years theory of political socialization (Jennings and Niemi 1981; Jennings 2007), which outlines how individuals develop the attitudes, values and habits that shape their political traits throughout adulthood during their ‘formative years’, i.e. the period of youth in which people are first developing political attitudes and values, and are particularly receptive to the influence of external factors. As individuals age, these characteristics ‘crystallize’ and become habitual, and the individual becomes less receptive to external influences, ultimately becoming less likely to change (barring a dramatic event that disrupts them, such as a war).

A group of individuals born around the same time, and who consequently experience similar socializing influences and contexts during their formative years, can be called a political generation (Mannheim 1928). As the context in which successive generations experience their formative years changes (and so they develop different political habits), lasting generational differences in political traits are created (Becker 1990). The literature identifies five broad theories that could account for such differences in Euroscepticism. The first reflects Down and Wilson’s (2013) argument that the level of EU integration during one’s formative years establishes the level of EU versus national government power one considers ‘normal’. The more powerful the EU during one’s formative years, the more likely one is to be comfortable with the EU’s influence and the less hostile they are towards it. Conversely, those who grew up witnessing the national government as supreme are likely to be hostile towards the loss of sovereignty resulting from EU integration. As the EU has only become more integrated over time, this theory expects older generations to have watched the EU move further away from the level of integration they are comfortable with and so are typically more Eurosceptic.

The second theory focusses on the consequences of social evolution for the development of political skills and values. As Western society has become increasingly economically and technologically developed, and as education has improved, new generations have been socialized into an environment in which they can take economic security for granted and can access more political information and skills (Inglehart and Welzel 2005). Consequently, they become more likely to develop post-materialistic traits, which emphasize political engagement and a political agenda in which issues such as individual freedom and autonomy are important (Inglehart and Welzel 2005). Their education also makes them more politically sophisticated, i.e. more skilled and capable of seeking out and utilizing political information (Dalton 2013). As successive generations become more post-materialist and politically sophisticated, they are expected to be less Eurosceptic, because: (i) they possess the political skills and motivation to understand the role of the EU in domestic politics, and so are less likely to find it threatening or confusing; (ii) they see the EU as an institution through which they can promote their post-materialist agenda; and (iii) their post-materialist values make them more supportive of international co-operation and institutions (Janssen 1991; Gabel 1998).

The third theory argues that nationalism is a powerful driver of Euroscepticism, particularly if it reflects an ‘exclusive’ identity (such as ‘English’) associated with the sovereignty, independence, and/or superiority of a particular nationality (McLaren 2002; Lubbers and Scheepers 2007; Ford and Goodwin 2014). As a multi-national, multi-cultural institution that weakens the direct power of nation-states, the EU is unappealing to people who hold such identities. Such individuals can also be more Eurosceptic because of their hostility towards migrants from other EU states. Younger generations are less likely to hold traditional conceptions of national identity or to share such nationalistic sentiments; consequently, they are expected to be more supportive of EU membership (Ford and Goodwin 2014).

Fourth is the economic capital theory, which argues that those with the skills, networks, and capital that allow them to take advantage of the economic opportunities of EU membership (such as working, investing, or trading in the single market), or that make them less likely to find themselves in competition with low skilled, poorly resourced European migrants, are more likely to support EU membership (Gabel 1998; Hooghe and Marks 2005). This theory could argue that younger generations should be more supportive of EU membership, because they have more human capital resulting from rising levels of education and training. It could also argue, however, that younger generations should be more Eurosceptic because of their increasing difficulties in achieving economic independence and security upon completing full-time education (Smets 2016).

Finally, the domestic politics theory argues that Euroscepticism reflects one’s loyalties to domestic political and social institutions. Those who identify with institutions that exhibit a Eurosceptic ideology (such as a Eurosceptic political party) or an association sceptical about the loss of national sovereignty (such as Protestant churches) are more likely to be Eurosceptic and to have the trait reinforced through their loyalty to and association with that institution (Nelsen, Guth, and Fraser 2001; Ford and Goodwin 2014; Scherer 2015). Younger generations are less likely to identify with many such institutions, in part because they are less likely to identify with many traditional social and political institutions overall, but also because of their tendency to be less Eurosceptic as outlined above (Dalton 2013). Consequently, they are less likely to exhibit or have reinforced attitudes that make them hostile towards EU membership.

Research design

APC analysis and political generations in the UK

This study uses APC analysis, a regression technique in which cohort, life-cycle and period effects are estimated simultaneously while controlling for traits relating to explanatory theories. In this instance, a cohort effect would reflect generational differences in Euroscepticism. Life-cycle or age effects reflect processes (biological, social or psychological) related to ageing, and period effects reflect contextual factors that affect everyone in a given population (Neundorf and Niemi 2014). The key challenge of APC analysis is the ‘identification problem’, i.e. the fact that in most survey data cohort, age and period are measured in the same unit: years (Neundorf and Niemi 2014). This means they are linear functions of each other and cannot be estimated in the same regression model.

Numerous techniques have been developed to enable at least tentative estimates of the three effects, the most appropriate of which for this study is the ‘coefficient constraint’ approach (Yang and Land 2013). This involves categorizing one of the variables and transforming it from an interval variable that shares linearity with the other two into a categorical variable that does not. The selection of the constraint can be problematic, as it cannot be identified from the data and is dependent on the judgement of the analyst, theory, and as much additional information as possible (Yang and Land 2013). A particularly useful technique for providing such information is generalized additive modelling (GAM): a semi-parametric regression method that allows the effect of the constrained variable to be estimated non-parametrically while still controlling for the other variables of interest (Grasso 2014; Keele 2008). This non-parametric effect can be visually compared with that in the standard regression model as a way of empirically supporting the chosen constraint.

As the interest of this study lies in generational differences in Euroscepticism, the constraint is applied to the cohort variable by categorizing respondents into political generations, which are defined on the basis of key stages in the social, economic, and political history of the post-war era that are expected to have had a substantial impact on the socialization of generations experiencing their formative years in that context. Numerous studies show that the economic, social, and political evolution of Western society is related to generational differences in a range of political traits associated with individual-level Euroscepticism, including political interest and sophistication, national identity, institutional identification and ideology (Van Deth and Elff 2000; Furlong and Cartmel 2012; Grasso 2014). Moreover, Down and Wilson (2013, 2017) argue that the historic development of the EU itself is vital in determining Euroscepticism, and they derive a generational classification to capture this effect which is very similar to that employed by these other studies.² There is a broad consensus, therefore, regarding the key stages of European social, economic, political, and technological evolution that have produced generational differences in both Euroscepticism and political traits related to it, which provides a good basis for the generational classification used in this study.

Where such research differs is in determining how individuals should be allocated to a generation. Down and Wilson’s (2013) analysis is problematic because they did so on the basis of when individuals ‘come of age’, i.e. turn 15, which assumes that people have developed the values that will shape their Euroscepticism by age 15, whereas political socialization literature suggests that the formative years begin at this age and can last until the mid-twenties (Jennings 2007). A more consistent approach with the impressionable years theory is to assign individuals on the basis of the period in which they spent the majority of this formative period (Grasso 2014). This study uses this approach to assign British citizens to one of six political generations derived from the research outlined above, as follows:

1. The Pre-War generation (born before 1925) experienced most of their formative years between 1908 and 1950, facing severe threats to physical and economic security from war and depression. These concerns dominated domestic politics, which was highly institutionalized and based primarily around mass membership bodies such as political parties (Grasso 2014). This period saw the first steps towards the formation of the European Coal and Steel Community (ECSC), an initiative broadly supported by people throughout much of Europe (though the UK did not join) because of the widely held view that it would prevent another war (Down and Wilson 2013). This generation is consequently expected to be broadly supportive of the EU.

2. The Post-War generation (born 1926–1945) experienced most of their formative years between 1941 and 1970. This period was dominated by post-war reconstruction and the establishment of the welfare state, prompting a dramatic rise in living standards and economic security (Dalton 2013). Life expectancy, for example, increased by nearly a decade between the early 1940s and 1970 (Hicks and Allen 1999). Institutions such as churches, unions, political parties, and social class continued to dominate social and political life, with millions members of parties and even more in trade unions (Grasso 2014; Keen, Apostolva, and Audickas 2016). This generation witnessed European integration move beyond the ECSC and the formation of the European Economic Community (EEC). Debates about the UK joining the EEC began and an application for membership was made in 1961. The fact that the UK was not a member of the EEC for much of this generation’s formative years, that they lack the experience of war that facilitated sympathy for the EEC among their predecessors, and that they are (relatively) uneducated, suggests that they should be more Eurosceptic than other generations.

3. The 60s–70s generation (born 1946–1957) experienced most of their formative years between 1961 and 1982 and were the primary beneficiaries of post-war reconstruction. They experienced unprecedented economic security, a well-funded welfare state and secure employment in low skilled but well-paid jobs: life expectancy rose above 70 for both men and women, and unemployment was consistently below 5% (Hicks and Allen 1999; Inglehart and Welzel 2005). Domestic politics was still largely shaped by traditional institutions and was increasingly ideologically polarized, but non-electoral participation was becoming more popular (Grasso 2014). One of the key experiences of this generation was to see the UK join the EEC in 1973. This, along with the greater economic security and capital of this generation compared with the Post-War, as well as their superior education and weakened attachments to domestic social and political institutions, should make them more supportive of EU membership than the Post-War generation.

4. The 80s generation (born 1958–1968) experienced most of their formative years between 1973 and 1993, in the period of economic and political crisis and reform accompanying the rise of the ‘new right’. Living standards and access to education continued to improve, with the proportion of 18/19-year-olds attending university more than doubling between the 1960s and 1970s alone (Mayhew, Deer, and Dua 2004), though the financial security and secure employment enjoyed by the 60s–70s generation was not as readily available (Grasso 2014). Like the 60–70s generation, the 80s generation saw the UK join the EEC during their formative years, but also witnessed considerable further integration and a dramatic increase in the salience of European integration in domestic politics; in Britain, for example, intense battles over Europe contributed to Margaret Thatcher’s downfall and tore apart the Major government. This generation is expected to be less Eurosceptic than their predecessors (except perhaps for the Pre-War generation) because despite the economic crises of the period they still had access to better living standards and more education, and there is little evidence of the erosion of ties between individuals and domestic political institutions slowing at this time. They also experienced a greater level of European integration during their formative years.

5. The 90s generation (born 1969–1981) experienced most of their formative years between 1984 and 2006. While this period also saw considerable economic uncertainty, they nonetheless grew up with greater living standards and education than their predecessors (both access to higher education and life expectancy continued to rise) (Hicks and Allen 1999; Mayhew, Deer, and Dua 2004). The 90s generation was the first for which all of their formative years occurred while the UK was an EEC/EC/EU member, and during which the Treaties of Amsterdam and Nice saw the EU gain considerably more control over domestic political issues, and led to the EU itself becoming a more salient domestic political concern (Down and Wilson 2013).

6. Finally, the Millennials were born after 1982 and experienced most of their formative years after 1997. Their access to education and training, as well as the integration of technology into their daily life, was unprecedented (Dalton 2013). Domestic politics was no longer shaped by ideological battles between Left and Right, but by conflict over which party could achieve commonly shared objectives – including EU membership – with all main parties broadly supporting the EU (aside from United Kingdom Independence Party (UKIP) from 1993) (Whiteley et al. 2013). Deterioration of links between younger citizens and traditional institutions such as political parties continued, as did the rising propensity of citizens to engage in informal political activity (Dalton 2013). This was accompanied by the rising popularity of alternative political parties, including the explicitly anti-EU British National Party (BNP) and UKIP, and strongly pro-EU Greens. The Millennials grew up in the most integrated EU, which had considerable influence over a range of domestic and foreign policy areas and yet more power for the European Parliament. This is expected to make them less Eurosceptic than previous generations.

Data and method

This study uses data from the British Election Study (BES) for 1979, 1997, 2005, 2010, and 2015. The BES is a nationally representative survey examining a wide range of political attitudes, including support for EU membership. The consistency of measures allows for direct comparison over time, and the range of data allows the theories of Euroscepticism outlined above to be studied. The BES does not use extensive measures of attitudes towards the EU, and most vary substantially from one survey to the next making them unsuitable for this analysis. It has consistently included, however, a measure that captures utilitarian assessments of whether EU membership is good or bad for the UK comparable to that in Eurobarometer and in previous studies (Gabel 1998; Down and Wilson 2013). In 1979, respondents were asked whether they felt being in the EEC made Britain better or worse off; in 1997, whether they felt the UK should continue as an EC member; and in 2005–2015, to indicate their approval for EU membership. These variables were recoded so that respondents who felt that EC membership made Britain worse off, that Britain should withdraw, or who disapproved of EU membership were identified as Eurosceptics.

Logit regression models predicted the probability of Euroscepticism with generation measured using dummy variables identifying the Millennials, 90s, 80s, 60–70s and Pre-War generations (the Post-War generation was the reference category). Life-cycle effects were modelled using respondents’ age, and period effects using dummy variables for each survey year except 1979 (the reference category).³ A control for gender was also included. The effect of post-materialism and education was measured by respondents’ interest in politics and highest educational qualification, following the approach used by Dalton (2013) in which post-materialists are expected to be more politically sophisticated and engaged.⁴ The effect of national identity was represented by hostility to immigration (controlling for ethnicity).⁵ Economic capital was represented by respondents’ tenure and occupational social class, with those in jobs associated with higher income and qualification requirements, and who owned their homes, assumed to have more capital. The influence of domestic political loyalties was measured through respondents’ party and social class identification and religious affiliation.

Control variables for the influence of domestic political judgements were also included, specifically whether respondents felt the government handled the economy well, and retrospective and prospective assessments of both national and personal finances. This accounts for the correlation between domestic political judgements and Euroscepticism, which can result from individuals who either do not believe the EU to be a particularly salient issue, or do not have clearly formed judgements about EU membership, projecting their assessments of other political institutions (such as the government) onto the EU when asked to reach a judgement about it (Franklin, Van der Eijk, and Marsh 1995).⁶ All models were estimated with robust standard errors and tested with a linktest to ensure that the logit function was appropriate.⁷ Each model was repeated using GAM to ensure that the constraint applied to the cohort variable (categorizing respondents into political generations) produced a reasonable representation of the cohort effect (the GAMs are presented and discussed in Appendix B).

Results

Table 1 presents the results of the basic APC model (including the cohort, age and period variables plus gender control) and a full model including all the variables outlined above.⁸ It reports regression coefficients and standard errors, and predicted probabilities of being Eurosceptic expressed as a percentage, which provides a standardized measure to enable easier interpretation and direct comparison between variables and models. To examine the contribution of each individual theory to explaining generational differences, Table 2 reports the predicted probabilities of each generation being Eurosceptic from models in which only variables relating to each theory were included, allowing comparison of the effect of the theories on generational differences in support for the EU. The full regression models from which these probabilities were calculated are reported in Appendix C.

Table 1. APC analysis of Euroscepticism in Britain, 1979–2015.

	Basic model		Prob (%)	Full model		Prob (%)
	Coef.	S. er.		Coef.	S. er.
Generation (Post-War)			38.5			32.3
Millennials	−0.51*	0.21	27.4	−0.33	0.31	25.6
90s	−0.43†	0.15	28.9	−0.14	0.23	29.4
80s	−0.21	0.11	33.7	0.07	0.16	33.9
60–70s	−0.08	0.07	36.7	0.09	0.11	34.4
Pre-War	−0.29†	0.10	31.9	−0.34*	0.14	25.0
Age	0.01‡	0.00		0.02†	0.01
Year (1979)
1997	−0.35‡	0.09		−0.14	0.14
2005	−0.47‡	0.11		−0.23	0.20
2010	−0.18	0.13		−0.03	0.19
2015	−0.39†	0.14		−0.08	0.22
Gender (female)	0.06	0.04		−0.02	0.06
Political interest (none/little)						37.1
Some				−0.25‡	0.07	31.3
Quite a lot				−0.32‡	0.07	29.9
A great deal				−0.40‡	0.10	28.4
Education (no qualification)						39.8
GCSE D-G/NVQ 1				−0.23*	0.11	34.4
GCSE A-C/NVQ 2				−0.25†	0.08	34.0
A-Level/NVQ 3				−0.38‡	0.09	31.1
Sub-degree HE/NVQ 4				−0.50‡	0.08	28.5
Degree/higher degree				−0.88‡	0.10	21.5
Tenure (rented)						35.2
Own home				−0.24‡	0.06	30.0
Social class (employer in large org./professional/managerial)						28.5
Intermediate/non-Manual				0.12	0.08	31.0
Small org. employer/manager				0.25*	0.10	33.9
Supervisory/skilled manual				0.25†	0.09	33.9
Semi/un-skilled manual				0.26†	0.09	34.2
Other/never worked				0.26*	0.12	34.0
Immigration (bad for Britain)						35.7
Good for Britain				−0.59‡	0.06	23.5
Ethnicity (White European)						31.8
Non-White European				−0.18	0.13	28.1
Class ID (none)						30.6
Working class				0.13*	0.06	33.5
Middle class				−0.02	0.07	30.2
Other/don’t know				0.13	0.15	33.5
Party ID (Conservative)						37.3
Labour				−0.47‡	0.07	27.2
Lib Dem				−0.55‡	0.09	25.5
SNP/Plaid Cymru				−0.34†	0.13	29.6
UKIP/BNP				1.34‡	0.20	69.4
None				−0.20*	0.09	32.7
Other				−0.27	0.16	30.7
Religion (none)						32.7
Church of England/Wales/Episcopal				−0.03	0.06	32.0
Roman Catholic				−0.22*	0.09	28.1
Presbyterian/Church of Scotland				−0.33†	0.11	26.0
Other Christian				−0.11	0.09	30.4
Other				0.19	0.16	36.9
Plus controls for domestic political judgements
Constant	−0.91‡	0.18		0.29	0.30
Obs	14538			8582
Prob > chi^2	0.000			0.000
Pseudo r^2	0.03			0.11

Note: Predicted probabilities calculated using Stata’s margin command, with other variables held at means.

Source: BES 1979, 1997, 2005, 2010, and 2015.

*Statistically significant at 95% confidence level.

†Significant at 99%.

‡Significant at 99.9%.

Table 2. Predicted probability of Euroscepticism (%).

	Basic	Cognitive mobilization and post-materialism	Human capital	National identity	Domestic politics	Full
Millennials	27.4	24.8	24.2	27.4	28.6	25.6
90s	28.9	28.8	28.5	29.3	29.2	29.4
80s	33.7	32.7	33.7	35.0	34.3	33.9
60s–70s	36.7	34.9	33.7	37.0	36.6	34.4
Post-War	38.5	34.9	38.1	37.8	36.4	32.3
Pre-War	31.9	29.1	30.7	31.2	28.9	25.0
Average Difference	6.5	7.3	8.7	6.7	4.5	5.4
Millennials vs. Post-War	−11.1	−10.1	−13.9	−10.4	−7.8	−6.4
Pseudo r^2	0.03	0.06	0.05	0.06	0.07	0.11

Source: BES 1979, 1997, 2005, 2010, and 2015.

The basic model confirms that there are indeed significant generational differences in British Euroscepticism, as well as life-cycle and period effects. The period coefficients show that Euroscepticism has fluctuated in the electorate since the 1970s, most likely in response to changes in the salience of EU membership and related issues (such as immigration). The age coefficients confirm Down and Wilson’s (2013) finding that people tend to become more Eurosceptic as they get older, reflecting the fact that people become less able to take advantage of many of the most salient opportunities offered by EU membership as they age. During youth, for example, people tend not to be married, have families, established careers or be rooted in communities, which means they are more able to take advantage of opportunities such as the freedom to travel, work, live, and study in other EU countries. Once people get older and become less transient, however, these opportunities become less relevant, and so they do not have the same reasons to view EU membership in such a positive light (Down and Wilson 2013).

As was also expected, this life-cycle effect is opposed by a generational decline in Euroscepticism. The most Eurosceptic generation is the Post-War, who typically have a greater than 38% likelihood of being Eurosceptic. From then, the likelihood of Euroscepticism falls and (up to the Millennials) by greater amounts with each generation.

There is essentially, therefore, a curvilinear generational decline in Euroscepticism, the steepest point of which occurs when the 90s generation, the first to spend all of their formative years inside the EU, enters the electorate. Figure 1 illustrates both this and the opposing life-cycle effect. It implies that the life-cycle effect is stronger than the generational effect, which the regression analysis confirms. For each year of age the likelihood of being Eurosceptic increases by almost 2%, compared with an average generational decline in the likelihood of being Eurosceptic of almost 3%. While every generation to enter the electorate since the Post-War has been less Eurosceptic, therefore, this is countered by the increase in hostility towards the EU throughout their lives. In light of the UK’s ageing population, this may help explain why the UK voted to leave the EU in 2016 despite the generational trend.

Figure 1. Age/generation and Euroscepticism in Britain. Source: BES 1979, 1997, 2005, 2010, and 2015.

Note: Values were averaged across age bands to provide a smoothed estimate of the proportion of Eurosceptic voters and make trends more reliable and easy to see.

The full model in Table 1 shows that all five theories help explain the generational differences. Virtually all of the coefficients have the expected effect: higher levels of political engagement and sophistication reduce Euroscepticism; the tenure and social class variables show that higher levels of financial security and resources make people more supportive of EU membership; hostility to immigrants has a strong and positive effect on hostility towards the EU; and domestic institutional identification has a strong effect, with those who identify with ‘hard’ Eurosceptic parties such as UKIP or the BNP being more likely to be hostile towards the EU than those who do not identify with a party at all, and substantially more likely than those who identify with the pro-EU Liberal Democrats. Those who identify as working class are more likely to be Eurosceptic than those with no class identity or who identify as middle class, and members of the Anglican or no church are more likely to be Eurosceptic than Catholics or Presbyterians. Given the geographic concentration of Presbyterians in Scotland, and members of the Anglican church in England, it is likely that this relationship reflects regional effects; however, Nelsen, Guth, and Fraser (2001) and WISERD (2016) have shown that the effect of religion on Euroscepticism persists even once many other factors (including region) are controlled for, suggesting that while the magnitude of the effect of religion may be over-estimated, it is likely to be important nonetheless.

Controlling for these theories helps explain the generational effects, though they do not fully account for the differences, and the overall contribution to explaining differences in Euroscepticism is limited, as the pseudo r² of 0.11 shows.

Table 2 shows that the domestic politics model makes a larger contribution to explaining the generational differences than the other theories, with the average difference between the Millennials’ likelihood of being Eurosceptic and that of the rest of the electorate being four and a half per cent, compared with six and a half per cent in the basic model and five and a half per cent in the full model. It also shows that by controlling for traits such as the Millennials’ weaker identification with political parties or other Eurosceptic social institutions, their probability of being Eurosceptic is greater than for the other models. The 2015 BES, for example, showed more than a third of Millennials did not identify with a social class, a quarter did not identify with a political party, and more than half had no religious affiliation; this compares with an average among the older generations of 27% with no social class identification, 13% with no party identification, and a third with no religion. The identity and cognitive mobilization theories are also notable: younger generations are less likely to be hostile to immigration (more than 40% of Millennials in the 2015 BES felt immigration was good for the UK, compared with one in three of older generations), and are typically more highly educated, than older generations (a quarter of Millennials, and a third of the 90s generation, had a degree, compared with one in ten of the Post-War generation). The remaining difference between the 90s and Millennial generations and their elders is likely to reflect the differences in prevailing levels of EU integration during their formative years.

Finally, the analyses also show that the Millennials in particular face distinctive pressures that make them more Eurosceptic than we might expect. First, while they are among the most highly educated in the electorate, they are also the most politically apathetic: half of Millennials had little or no interest in politics in the 2015 BES. This makes them less likely to be informed about the role of the EU in domestic politics and so more hostile towards it, which explains why the average difference between the Millennials’ likelihood of being Eurosceptic and that of their elders is larger for the cognitive mobilization model in Table 2 than might be expected. Even more substantial is the effect of the Millennials’ lack of economic capital; the average difference between the Millennials’ chance of being Eurosceptic and that of their elders was greater for the human capital model than all others, and the difference between them and the other younger and more pro-EU generations (the 80s and 90s) increased relative to the basic model. This reflects the influence of the Millennials’ relative economic precocity, which makes them less able to exploit the economic advantages of EU membership. In the 2015 BES, more than 60% of Millennials were living in rented accommodation, while more than half of the 90s generation owned their homes in the 1997 BES.

Conclusion

This paper has examined the generational divide in Euroscepticism that was so clearly illustrated in the UK’s recent referendum on EU membership. Following Down and Wilson (2013, 2017), it has utilized APC analysis to disentangle the effects of cohort, life cycle, and historic context, and assessed the extent to which differences in the formative experiences of successive generations have led to substantial age differences in support for EU membership. This study has also gone further in examining the characteristics that could explain the generational decline, by considering the role of traits associated with individual-level differences in Euroscepticism in previous research. It has provided, therefore, the most detailed assessment of the extent and causes of generational differences in Euroscepticism to date.

The analyses confirm that there is a generational decay to Euroscepticism, which is caused by a combination of factors reflecting the formative experiences of successive generations, including: the increasingly integrated relationship between the EU and member states; rising levels of economic security; improvements in access to education; the weakening tendency to exhibit exclusive nationalist sentiments; increases in human and economic capital; and the declining tendency to identify with Eurosceptic (overtly or otherwise) domestic social and political institutions. Most of these changes have occurred somewhat consistently over time, which is why there is an almost linear generational decline in Euroscepticism from the Post-War generation. That said, it is notable that rising levels of political engagement and of economic capital have not been linear, and this has weakened the magnitude of the generational decline. In particular, the Millennials’ unusually low levels of political interest, and the distinctive challenges they face in becoming economically independent, led to Euroscepticism not being as low in the youngest generation as we might otherwise expect.

The analyses also showed that this generational trend is opposed by a life-cycle effect in which Euroscepticism increases with age. This is suggested to reflect the impact of changing life circumstances as people age on their capacity to take direct advantage of EU membership. This effect is stronger than the generational effect, meaning that while each successive generation has entered the electorate being less Eurosceptic than their elders, by the time they retire they are likely to be more Eurosceptic than the previous generation was when they first entered the electorate. A final point of note is that for the wide range of traits accounted for in the APC models, even the most successful in terms of predictive power only achieved a pseudo r² of 0.11; the basic APC model figure was just 0.03. In short, this shows that even with all of the established theories of Euroscepticism accounted for, along with the influence of age, period, and cohort, their collective contribution to explaining Euroscepticism is limited. Moreover, there has been little research into the nature of the relationship between the life cycle and Euroscepticism that could sustain a more detailed and comprehensive explanation for the effects identified above. While all of the above theories are important in explaining why a given individual may be more or less Eurosceptic than another, therefore, there remains considerable scope for future research to identify other, potentially even more important, traits that can account for differences in Euroscepticism both between generations and throughout the life cycle. A further requirement for future research is to explore the generalizability of these conclusions to other member states. This analysis was focussed on Britain because of the limitations in available multi-national data sets. While previous studies suggest that Euroscepticism follows broadly similar trends and relationships across EU member states, there is nonetheless scope for the conclusions of this research to be tested in other national contexts.

Finally, this research also helps explain the result of the UK’s referendum on EU membership, and sheds light on potential future referenda elsewhere. The data does not imply that a majority of voters felt EU membership was bad for the UK as recently as the 2015 general election, even though a majority voted to leave in 2016. This suggests either that anti-EU sentiment became far more widespread between the 2015 election and 2016 referendum, that pro-EU sentiment is over-represented in the BES, or that the referendum result was not solely a reflection of Euroscepticism but of other attitudes as well. All are possible, and are likely to be true to some extent, though determining this is beyond the scope of this research. In any event, it is unlikely that the majority of UK citizens who voted to leave the EU did so solely because they believed EU membership was bad for the country, but because of other concerns as well relating to their national identity, political engagement, associations with other political institutions, and judgements about the performance of the national government.

Acknowledgements

Earlier versions of this paper were presented at the Annual Conference of the Elections, Public Opinion and Parties specialist group of the Political Studies Association in 2016, and the 2016 conference 'The Impact of the 2016 Referendum on UK Membership of the EU' at Newcastle University. The authors would like to thank participants at both events for their useful questions and feedback, as well as the journal editors and anonymous reviewers for their comments and feedback, all of which greatly improved the research. Any errors remain the responsibility of the authors alone.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Stuart Fox is a Quantitative Research Associate in the Wales Institute of Social and Economic Research, Data and Methods. His PhD thesis examined generational trends in political apathy and alienation in the UK, and considered the effects on the political behaviour of young people. His current research focusses on the political engagement, participation and characteristics of young people in Britain. He has also participated in research regarding the effect of religious identification on political attitudes, and on the consequences of higher education for political and civic engagement.

Sioned Pearce is a research associate in the Wales Institute of Social and Economic Research, Data and Methods. She has conducted research in a range of environments including academics and the third-sector, examining devolution and deprivation, young people and political engagement, and the impact of academic research on policy and civil society. Her current priorities include the study of young people's political engagement and participation, and of 'impact' as a contested concept in understanding the relations between state and civil society in the context of devolution and constitutional change.

Additional information

Funding

This work was supported by Economic and Social Research Council [grant number ES/L009099/1].

Notes

1. Around 61% of under-30s voted to remain in the EU compared with a third of over-65s (WISERD 2016).

2. Grasso (2014) identified five political generations in Western European electorates, including Pre-WWII (born 1909–1925); Post-WWII (1926–1945); 60–70s (1946–1957); 80s (1958–1968); and 90s (1969–1981). Van Deth and Elff (2000) identified the Pre-War generation (1910–1930), silent generation (1931–1940), protest generation (1941–1955), lost generation (1956–1970), and pragmatic generation (born after 1970). Down and Wilson (2013) identified six generations in relation to the key stages of European integration. The first was born before 1939, meaning that they came of age (i.e. turned fifteen) before integration began in 1952; the second (1939–1952) coming of age between the Treaty of Paris and the Merger Treaty; the third (1953–1972) between the Merger Treaty and the Single European Act; the fourth (1973–1978), between the Single European Act and the Maastricht Treaty; the fifth (1979–1987) between Maastricht and the introduction of the euro; and the sixth being born after 1987.

3. The use of a quadratic age function was also tested but rejected.

4. The analyses were unable to distinguish respondents who had completed full-timed education as this was not available in the 1997 and 2015 BES, meaning that while for some respondents their highest qualification is unlikely to change for the rest of their lives, for others (particularly the young), it could change as they complete higher education or further training. As a result, the effect of education is likely to be under-estimated.

5. While a direct measure of nationalist sentiment would be preferable, none was available and so a proxy was used on the assumption (verified in Appendix A) that those who are hostile to immigrants are more likely to hold exclusive nationalist sentiments.

6. Identifying with UKIP and the BNP were treated identically, reflecting their shared “hard” Euroscepticism.

7. Different ways of operationalizing each theory were assessed on the basis of theoretical logic, model fit, model parsimony and the significance of effects before the forms presented above were determined. Details are not presented because of space limitations but are available on request. The tests included interaction effects between generation and education and political interest (cognitive mobilization and post-materialism model), support for immigration and generation (identity model), and generation and party identification (domestic politics model).

8. The full range of regression models are available in Appendix C.