Explaining voting in the UK's 2016 EU referendum: Values, attitudes to immigration, European identity and political trust.

We consider the role of values as determinants of voting in the UK's 2016 EU referendum. First, we consider and clarify 'values' conceptually, before proceeding to utilise Schwartz's basic human values as the most appropriate. Second, we theorise how these basic human values determine both voting in the referendum as well as three of the most consistently demonstrated determinants of voting in that referendum: attitudes to immigration, identifying as European and trust in politicians. Finally, we demonstrate that this psychological theoretical framework effectively predicts voting behaviour in the referendum using multigroup structural equation modelling. Overall, we show that voting dynamics in the referendum are likely to reflect deeper, more personal psychological predispositions than those captured by existing explanations. In doing so, we make an original contribution to the literature on political psychology related to Brexit, Euroscepticism, political attitudes and electoral behaviour.


Introduction
On June 23, 2016, the government of the United Kingdom held a national referendum to measure public opinion on the country's continued membership in the European Union, with an explicit promise that the outcome would set the government's future policy following decades of Eurosceptic agitation on the matter. The final result saw 17,410,742 votes for 'Leave' (51.9%) and 16,141,241 votes for 'Remain' (48.1%) in a surprisingly high turnout of 72.2% of registered voters. The outcome confounded the predictions of polls, academics and financial markets, resulted in a radical departure from 43 years of British involvement in the "European project" and represented a "leap into the unknown" for one of the EU's "Big Three" that has yet to be resolved. For the EU, it represented an additional crisis and its biggest vote of no confidence to date.
There have since been numerous scholarly explanations for voting in this referendum. Of these explanations, the role and predictive power of 'values' has been implied and discussed by academics, not only regarding the EU referendum but also in other elections in which so-called populist parties and candidates have performed well. As we show, however, in these works, values have been conceptually underspecified and, instead, operationalised with proxies such as attitudes to social, non-economic issues. Moreover, despite considerable discussion, there have also been few attempts to empirically test the claim that variation in values were a key determinant of variation in voting in the referendum.
The motivation of this article is to contribute to the literature explaining voting patterns in the 'Brexit' referendum in three ways. First, we conceptually clarify 'values' as being personal transsituational goals, varying in importance, that serve as guiding principles in the life of individuals (Schwartz, 1994: 21). As such, they are motivations for personal life choices and distinct from political attitudes, although they can and do predict situational including political, attitudes and behaviour. They are expected to be highly stable over the life course. We consider a number of established values schema from the psychology literature, before explaining why we consider Schwartz's theory of basic human values is the most appropriate for explaining voting in the UK's 2016 EU referendum. Second, we theorise how these basic human values determine both voting in the referendum and well as three of the most consistently demonstrated determinants of voting in that referendum: attitudes to immigration, identifying as European and trust in politicians. Finally, we demonstrate that this psychological theoretical framework effectively predicts voting behaviour in the referendum using multigroup structural equation modelling. Overall, we show that voting dynamics in the referendum are likely to reflect deeper, more personal psychological predispositions than those captured by existing explanations. In doing so, we make an original contribution to the literature on political psychology, related to Brexit, Euroscepticism, political attitudes and electoral behaviour.

Existing explanations for voting in the EU referendum
The referendum has resulted in a large amount of scholarship, with one of the most well-developed areas being explanations of variation in individual-level voting behaviour. In this section we identify roughly six groups of such explanations: identity, policy attitudes, 'populism', economics, socio-demographics and values.
First, the extent to which an individual identifies as a European has been shown to have increased their probability of voting 'Remain' in the referendum (Carl et al., 2019;Curtice, 2017: 21;van de Vyver et al., 2018). European identity has been defined in different ways but is perhaps best thought of as the extent to which an individual sees 'Europe' and being 'European' as an essential part of his/herself and as something to which one is emotionally attached (Carey, 2002;Hooghe and Marks, 2005;McLaren, 2004). The concept is built upon 'social identity theory', which argues that identity is rooted in evolution and results in a strong desire to defend the territory that aligns with one's identity, while it has also been theorised to lead to more practical preferences regarding transfers of competences, solidarity with other Europeans and the EU's redistributive policies (Carl et al., 2019). 1 Second, the importance of attitudes to immigration in the Brexit vote has also been evidenced by several studies (Goodwin and Milazzo, 2017;Arnorsson and Zoega, 2018;Swales, 2016). Moreover, an individual's attitudes to immigration have also long been shown as a key determinant of support for European integration (De Vriese and Boomgaarden, 2005;McLaren, 2002;Sniderman et al., 2004;Heath and Tilley, 2005) with the theoretical argument being that those with anti-immigration views are likely to oppose European integration because the EU's legal regime enshrines the right to migrate and historically prioritises economic and social liberalism in policy-making, both of which may encourage migration.
Third, scholars have emphasised the role of anti-establishment feelings and distrust in politicians in causing individuals to vote to Leave (Hobolt, 2016;Goodwin and Heath, 2016a). Again, these attitudes had already previously been shown to predict Euroscepticism and voting against European integration in referendums (Treib, 2014). Krouwel and Abts (2007) argue that Euroscepticism is the result of, first, broader declining trust in democratic institutions and, second, mobilisation of political discontent by populist entrepreneurs who criticise the entire political elite which, in most member states, tends to favour European integration. According to this explanation, anti-Europeanism results both from lower trust in politicians to pursue European integration for benevolent ends (with trust being particularly important given the arguably distant, complex and technocratic nature of the EU) and from a motivation to punish mainstream political actors for issues unrelated to the EU (see also Harmsen, 2010).
Fourth, scholars have pointed to economic anxieties as a motivating factor behind voting for Leave, either in terms of a backlash against fiscal austerity and inequality (Dorling, 2016) or as a protest by the 'losers of globalisation' (Hobolt, 2016), often defined as historical distinct-level prevalence of manufacturing jobs (Becker et al., 2017). However, others have claimed that economic variables did not play a role when controlling for attitudinal variables at the individual level (Matti and Zhou, 2017;Kaufmann, 2016;Inglehart and Norris, 2016).
Fifth, and related, various socio-demographic variables have been linked with voting in the referendum, both at the individual-level but, particularly in the months immediately following the referendum, at the district-level. These include country; race; gender; urbanity; country of birth, religion and education. Residing in Scotland, Northern Ireland and London have been shown to have increased one's chances of voting for Remain, vis-à-vis living in England and Wales (e.g. Heath, 2016a, 2016b;Arnorsson and Zoega, 2018). White ethnicity, being male, living outside of major cities, being native born, having low education and being Protestant have all been associated with an increased chance of voting Leave. Collectively, some have characterised those fulfilling a number of the above socio-demographic and economic anxiety predictors of voting Leave as being 'left behind' voters Heath, 2016a, 2016b).
Sixth, and finally, 'values' have increasing been pinpointed as the key predictor of voting in the referendum. Some scholars have operationalised 'personal values' as political attitudes towards social issues such as the death penalty and immigration (Kaufmann, 2016;Arnorsson;Zoega, 2016;Andreouli et al., 2019) while others have labelled this division between social liberals and social conservatives as a value divide (Goodwin and Heath, 2016a;Ford and Goodwin, 2017;Evans and Anand, 2017). Values have also been argued to be the basis of a culture war of which the 2016 EU referendum was just one expression (Koch, 2017). Still others have argued that Brexit was a manifestation of a value divide between those supporting multiculturalism and cosmopolitanism, on the one hand, and 'populist' values, on the other (Inglehart and Norris, 2016). Furthermore, Hanel and Wolf (2019) demonstrated differences in the mean value scores of Schwartz' basic human values between Remain voters and Leave voters.
We now turn to explaining why we see taking a values-based approach to explain voting in the referendum, as the above studies have done, as highly promising, both in terms of the theoretical premise that fundamental, personal values underpin political attitudes, which then affects behaviour, and practically both in terms of their explanatory and predictive potential. However, we also suggest that, thus far, values have not been sufficiently conceptually clarified, instead being muddled with political attitudes, nor based on an existing, theoretically sound values schema.

Values
Throughout the twentieth century, psychologists made numerous attempts to classify personal 'values'. The commonalities that tend to underpin each of these classifications are that their constituent 'values' are identifiable, are drawn from a finite set, tend to relate to each other in some systematic manner, vary little within individuals in the short term, vary more significantly between individuals and can be successfully used as predictors for attitudes on more specific, temporal issues and human behaviours. Brosch and Sander (2013: 3) define values as 'stable motivational constructs or beliefs about desirable end states that transcend specific situations and guide the selection or evaluation of behaviours and events.' Indeed, the importance of values as predictors of human activity was noted at least as early as 1961 by Allport, who stated 'personal values are the dominating force in life, and all of a person's activity is directed toward the realization of his values. And so the focus for understanding is the other's value-orientation-or, we might say, his philosophy of life (Allport, 1961: 543).' Some of the more prominent human value theories include those of Murray (1938), Rokeach (1973), Feather and Peay (1975, Maloney and Katz (1976), Hofstede (1980), Wicker et al. (1984), Cawley, Martin and Johnson (2000, Peterson and Seligman (2004), Schwartz (1992Schwartz ( , 1994Schwartz ( , 2012 and Talevich et al. (2017).
The use of 'values' to predict voting in the UK's 2016 EU referendum has thus far been inadequate because its use of non-economic political attitudes as values does not capture the defining features of values: their stability over the life course, their fundamentally personal, motivational characteristic that makes them predictors for both political and non-political attitudes and behaviours, and their relation to each other as part of a theoretically sound and empirically validated schema. Whereas values are abstract beliefs with different motivations underlying them, behavioural choices or attitudes are much more specific (Ajzen, 1993(Ajzen, , 2005Fishbein and Ajzen, 2015;Hitlin and Piliavin, 2004). There is only a limited number of values, yet behavioural choices and objects toward which individuals may form certain attitudes are countless. Furthermore, values are formed early in life and remain rather stable later on (Hitlin and Piliavin, 2004), while attitudes or behaviour are more susceptible to change. These considerations lead us to expect that values may operate as a driving force behind specific behavioural choices or attitudes (Davidov et al., 2008) The general mechanism between values, on the one hand, and attitudes and behaviours, on the other, is that individuals will choose to think and act in accordance with that which best promotes the realization of the values that they prioritise (Sagiv and Schwartz, 1995). In this article, we choose to use perhaps the most eminent and broadly utilised of values schema, Schwartz's theory of basic personal values (1992), which has been effectively used to predict attitudes and behaviours that are closely related to the referendum, as we discuss below, and, unlike other values systems such as Inglehart (1990), is primarily used at the individual-level as opposed to the country-level. Schwartz (1992) developed the theory to present a comprehensive framework to understand how universal individual values evolve, develop and operate. According to the theory, values are ordered hierarchically, and although individuals share the same values, they differ in the priorities they give to different values (see also Inglehart, 1990;Rokeach, 1973). Schwartz (1992) argued in his theory that there are ten important universal values with different motivations underlying them: security, conformity, tradition, benevolence, universalism, self-direction, stimulation, hedonism, achievement and power.
Furthermore, Schwartz suggested that the ten values could be organised by two higher-order dimensions (see Fig. 1). Universalism and benevolence values reflect the higher-order dimension of self-transcendence values and oppose power and achievement values which reflect the self-enhancement dimension. Stimulation and self-direction values reflect the higher-order dimension of openness to change and oppose security, conformity and tradition values that reflect conservation (Schwartz, 1992; see also Schwartz et al., 2012). Hedonism lies in between the dimensions self-enhancement and openness to change. Schwartz suggested that researchers may use more broadly defined or more fine-tuned values depending on their research question and explanation (Schwartz, 2012: 665).
Previous studies have shown that universalist individuals are more positive and individuals valuing the three conservation values are more negative toward immigration (e.g., Davidov et al., 2008;Davidov and Meuleman, 2012;Davidov et al., 2019;Sagiv and Schwartz, 1995;Beierlein et al., 2016;Schwartz, 2007), already shown to be a predictor of voting Leave. Moreover, Schwartz et al. (2010: 422) show that Schwartz's ten comprehensive personal values act as effective predictors of ten core political values (e.g. law and order, civil liberties etc.) and, ultimately, party choice at the ballot box. Furthermore, Schmidt et al. (2019) use Schwartz's basic human values to effectively predict voting in the UK's 2017 general election, while, as mentioned, Hanel and Wolf (2019) showed differences in the mean value scores between Remain voters and Leave voters using Schwartz's values.
We now explain why, of Schwartz's ten basic human values and four higher-order dimensions, universalism and conservation values may be particularly relevant for explaining voting in the EU referendum. To be clear, we are not making any claims about the validity of each of these perceptions about the nature of the UK's membership of the EU, but instead considering the value-basis of popular perceptions of the EU, as used by political actors during the referendum campaign and prior. The basic motivational goal of universalism is an understanding, appreciation, tolerance, and protection for the welfare of all people and for nature. Membership of the European Union is likely to be a specific goal resulting from valuing universalism because it prioritises the welfare of nations and citizens beyond the United Kingdom. To these individuals, the diversity of member states is likely to be seen as a positive good while the Union's institutional infrastructure is likely to be seen as an opportunity to break down barriers and reduce intolerance and increase understanding between peoples, resolve collective action problems for the benefit of all Europeans and protect the welfare of vulnerable non-British states and peoples from both internal and external oppressors. Furthermore, the EU transnational nature is likely to appeal to those seeking to protect the environment both in the UK and beyond, with which European law and treaties deal heavily.
Each of the three values comprising 'conservation' are likely to increase one's chance of voting Leave. Although the UK had taken part in European integration for over 40 years at the time of the referendum, the dynamic nature of integration and the institutions' increased power, following successive treaties, mean that the current European Union does not form part of popular perceptions of a traditional component of the British constitution, given the latter's considerably longer history (Grant, 2008). Moreover, European integration is argued to be alien to traditions of the British political and legal system, in terms of Parliamentary sovereignty, the majoritarian Westminster system, an uncodified constitution and the use of common law. Furthermore, Britons valuing tradition may be especially antipathetic to European Union membership because of the UK's relative historic isolation from developments on 'the continent', which has popularly been viewed as remote, distinct and, to some extent, singular (ibid), as well as unusually close traditional connections with some of its former non-European colonies, for example, sharing a head of state, and, finally, its Second World War experience, which did not end with its traditional institutions being discredited by conquest or defeat (Carl et al., 2019). Those valuing conformity are likely to be concerned by the increased societal heterogeneity that the EU legal regime fosters and which encourages change rather than conformity towards old national cultural norms. Those valuing security may similarly be concerned by the perceived destabilising effects of European Union membership, increased societal heterogeneity and, increased societal heterogeneity and, as a result, an increased perceived risk of societal disorder Legally, they may also be concerned that the UK could be bound to pursue policies against its will or contrary to its perceived interests.
We also expect universalism and conservation to be determinants of three of the key attitudinal determinants of voting in the referendum as already identified by the literature. Following the logic of the discussion above, those valuing universalism are more likely to embrace a European identity, with the opposite true for those holding conservation values. Valuing universalism has already been shown to positively affect attitudes to immigration with, again, the opposite true for those who value conservation. Finally, we also expect values to relate to trust in national politicians. National politicians explicitly represent the nation, they are expected to conform to the norms of the country's political system and traditions and to protect the nation state from external influences, including those which may stem from the European Union. By taking part in extant institutions and holding offices, they uphold and represent the political status quo and, to some extent, social hierarchy. Thus, their acts may corroborate particularly with the motivations underlying individuals who attribute high importance to conservation values (Zhai, 2018). By way of contrast, universalistic individuals may be less concerned with local interests, less trustful and respectful of the UK's national political traditions and rather occupied with more global interests such as those of the EU.

Hypotheses
The above discussion of previous explanations for voting in the EU referendum and of values lead us to the following hypotheses.
• (H1) European identity increases one's chance of voting to remain in the EU. • (H2) Support for immigration increases one's chance of voting to remain in the EU. These relationships are outlined in Fig. 2.

Data and measures
Our data source is the 8th round of the European Social Survey (2016) collected in the UK between September 1, 2016 and March 20, 2017. The ESS employs strict probability samples of the population aged 15 and older, and the total sample size in the UK in this round was N = 1,959.
In this round, the ESS asked respondents in the UK if and how they voted in the 2016 referendum on EU membership. Of the 1,959 Britons that the ESS surveyed in the UK, 703 (35.9%) responded that they voted to remain a member of the EU, 712 (36.4%) responded that they voted to leave, 509 (26%) responded that they did not vote, and 35 (1.8%) responded that they either spoiled their ballot, did not know how they voted or refused to answer. These percentages are largely in line with the actual referendum result. Thus, we created a dichotomous variable-Remain-that is coded as 1 for those who voted to remain in the EU and 0 for those who voted to leave the EU. The variable contains the voting decisions of (1,959-544 = ) 1,417 individuals.
Our basic human value variables derived from the ESS 21-item battery that measured the ten human values (with two items for each value, except universalism which was measured with three items). Survey participations were asked to respond to the following statements: 'Now I will briefly describe some people. Please listen to each description and tell me how much each person is or is not like you'. Twenty-one statements were then described, the somewhat lengthy details of which can be found in the questionnaire (European Social Survey, 2016). Possible responses were ordinal, ranging from 1 (very much like me) to 6 (not like me at all). We reversed the ordering of the responses so that high numbers indicated a high importance of the values. Universalism was operationalised as a latent variable measured by three items inquiring about the importance of equal treatment and opportunities for all, listening to and understanding different people and caring for nature and looking after the environment. Conservation was operationalised as a latent variable measured directly by the items of the three values tradition, conformity and security. The two tradition items inquired about the importance of modesty and tradition, the two conformity items asked about the importance of following rules and behaving properly, and the two security items reported the importance of living in secure surroundings and having a strong state which defends its citizens. The standardised factor loadings of the value items were acceptable and ranged between 0.38 and 0.71 (Brown, 2015).
European identity was measured by the question 'how emotionally attached do you feel to Europe (Europe in general, not the European Union)?' Participants responded on a scale ranging from 0 (not at all emotionally attached) to 10 (very emotionally attached). Opposition to immigration was operationalised as a latent variable measured by three questions: 'to what extent do you think [country] should allow people of the same race or ethnic group as most [country]'s people to come and live here?'; 'How about people of a different race or ethnic group from most [country] people?'; and 'How about people from the poorer countries outside Europe?' The questions had four possible responses ranging from 1 (allow many) to 4 (allow none). The standardised factor loadings of the opposition to immigration items were high and ranged between 0.84 and 0.96. Finally, respondents provided their score for trust in politicians in response to the statement 'tell me on a score of 0-10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust'. Table 1 lists the constructs, the items measuring them and their response categories. It is worth noting that exactly 50% of the respondents voted to remain in the EU. Table 1 also lists the socio-demographic variables that we used as control variables in our model. Age was measured in years. Gender was a dichotomous variable (0 = female, 1 = male). Education measures whether a respondent had secondary or tertiary education or none of these. Perceived income was measured on a 4-point scale ranging from 1 (living comfortably on present income) to 4 (finding it very difficult on present income). 2 Urban-rural was a dichotomous variable (0 = living in a big city, suburbs, or town/small city, 1 = living in country village, farm or home in the countryside). Moreover, minority, native born and protestant were dichotomous variables (0 = no, 1 = yes). Finally, based on NUTS nomenclature geographic unit dummies were used to indicate whether a respondent resides in Scotland, Wales, Northern Ireland or London. Previous studies have shown that older people living in rural areas in the UK with a lower level of education are more likely to oppose the EU (see, e.g. Alabrese et al., 2019, for individual-level socio-demographic associations with vote choice in the 2016 referendum).

Method
To test the hypotheses presented above in Fig. 2, we first estimated bivariate structural equation modelling (SEM) models of each of Schwartz's ten basic values predicting voting to remain to show that indeed only universalism and conservation values are in fact associated with voting behaviour in the EU referendum. Next, we performed similar bivariate SEM models between European identity, opposition to immigration, and trust in politicians, and voting behaviour. Finally, we estimated a full structural equation model to examine whether our two human values-universalism and conservation-predicted voting to remain both with direct effects and via their effect on the three endogenous variables European identity, opposition to immigration and trust in politicians while controlling for the socio-demographic variables listed above. We used the software Mplus version 8 (Muthén & Muthén, 1998 and weighted least squares methodology (WLSMV; Muthén et al., 1997) to predict the binary outcome voting behaviour with probit coefficients. The treatment of missing data in WLSMV estimation is based on pairwise deletion (limited information) (Asparouhov and Muthén, 2010).

Results
In Table 2, we present the coefficients of the bivariate SEM models of all ten basic human values when used to predict voting to remain. 3 As predicted by our hypotheses, universalism is associated with an increased likelihood of voting to remain in the EU, while conservation values have the opposite association. The associations of voting to remain in the EU with the other values are considerably lower or insignificant. 4 In Table 3 we again present coefficients from respective bivariate SEM models predicting voting to remain in the EU. 5 The effect of each variable added here-European identity, opposition to immigration, trust in politicians-is in the expected direction. Notably, the effects of European identity and opposition to immigration are of a considerably larger magnitude than those of the other predictors.
In the next step, we calculated a structural equation model testing our full theoretical framework including the direct and indirect effects as outlined in Fig. 1. Due to missing data, 50 cases were excluded from this model. The global fit measures were satisfactory (CFI = 0.901; RMSEA = 0.033) (Hu and Bentler, 1999;Marsh et al., 2004). Table 4 presents the standardised coefficients. Starting with the most proximal relationships, we can see that European identity and opposition to immigration had statistically significant effects in the expected direction on voting to remain in the EU. Thus, identifying with Europe and favourable attitudes toward immigration resulted in a higher probability to vote Remain, supporting Hypotheses 1 and 2. In contrast to the expectation postulated in Hypothesis 3, there was no evidence of an effect of trust in politicians on voting to remain. The direct effects of the conservation and universalism values on voting to remain were both in the expected direction and statistically significant but of a lower magnitude than European identity and opposition to immigration, providing support for Hypothesis 4 and 5. By way of contrast, some of the effects of the two values were large when predicting the three endogenous variables, European identity, opposition to immigration and trust in politicians. Universalistic individuals were more likely to identify with Europe and to favour immigration. By way of contrast, conservative individuals were less likely to favour immigration and more likely to trust politicians, supporting Hypotheses 6, 8, 9, and 11. Hypothesis 10-predicting a negative effect of universalism on trust in politicians was not supported by the data. Hypothesis 7-predicting a negative effect of conservation on European identity-was supported by the data, although the standardised effect was weak.
Some of the control variables showed noteworthy effects. Voting remain in the EU was positively related to education, belonging to an ethnic minority, protestant denomination, and residing in Scotland. Moreover, European identity was positively related to education and negatively to age, perceived income satisfaction, gender (males are less attached to Europe) and being born in the UK. Opposition to immigration was positively related to age, perceived income satisfaction and being born in the UK and negatively to education and residing in Northern Ireland. Trust in politicians was positively related to education and protestant denomination and negatively to perceived income satisfaction and being born in the UK. Conservation was positively related to age, protestant denomination and residing in Northern Ireland and negatively to education and being born in the UK. Universalism was positively related to age, education and residing in London.
In Fig. 3, we show these standardised effects within our theoretical framework, as shown in Fig. 2. As we can see, the strongest effects, in descending order, are: the negative effect of universalism on opposition to immigration; the positive effect of conservation values on opposition to immigration; the positive effect of European identity on voting remain; the positive effect of universalism on European identity; and the negative effect of opposition to immigration on voting Remain.
Finally, we examined whether the indirect effects of basic values were significant and in the expected direction. The results of the mediation analysis are reported in Table 5. The table demonstrates that both the indirect and total effects of the two values were significant and in the expected direction. The results suggested that the mediation is stronger for universalistic values. While the direct effect of universalism on voting Remain was weak, the indirect influence via (increased) identification with Europe and favourableness towards immigration was considerably larger. By way of contrast, the influence of conservative values on voting Remain was mainly direct and only to a smaller extent mediated via (decreased) identification with Europe and opposition to immigration. Table 5 also demonstrates that the total effects of values were higher than the direct effects of the more proximal predictors of Remain, European identity and attitudes towards immigration, underscoring the importance of values as direct and indirect predictors of voting to remain in the EU.
To test the robustness of our findings we estimated a model using party support as an additional mediator distinguishing between parties officially supporting remaining in the EU (all parties except the UK Independence Party (UKIP), the Democratic Unionist Party (DUP) and the Conservatives) and those supporting leaving the EU (UKIP and the DUP) with those officially neutral-the Conservative-towards the issue as the reference category). The effects of the two human values, European identity, opposition to immigration and trust in politicians were essentially the same. As an additional robustness test, we re-estimated our full model (without party support) and used net household income instead of perceived income satisfaction. This reduced the sample by 287 cases. None of the income effects was significant. All other results remained the same. Finally, we re-estimated the full model (without party support and with perceived income satisfaction) using the full information maximum likelihood procedure that assumes that missing values were missing at random (Schafer and Graham, 2002). The results were once again essentially the same.

Summary and discussion
In June 2016, UK voters decided by a small majority in a national referendum to leave the EU. This result was not only surprising but has also confronted the UK and the EU with an unprecedented situation in which a large member of the EU has chosen to exit the union after several decades of membership. As "Brexit" is likely to leave the UK and the rest of the EU in a position of uncertainty with numerous unknown consequences, it is highly relevant to understand how this voting behaviour came about. There have since been several explanations for individual-level voting patterns in the referendum. Of these explanations, the role and predictive power of values has been underlined. However, these studies used proxies for values, and none of them used a theory of basic human values systematically to explain this voting behaviour. In this paper, we tried to fill the gap by explaining the voting dynamics in the UK's 2016 referendum using basic human values as general motivational life goals that guide behaviour. We considered other more proximal predictors which have been shown to be relevant for the UK voting, namely, European identity, opposition to immigration and political trust. For our analysis we used the 8th round of the ESS data from 2016, collected just shortly after the referendum took place.
Overall, our analyses support our initial theoretical proposition that basic human values influence citizens' electoral behaviour regarding their country's membership in the European Union. In particular, we produced and tested a theoretical framework whereby two values-conservation and universalism-directly (at least partly) affected voting in the referendum and also did so indirectly via their effects on the three more proximal attitudinal factors, European identity, opposition to immigration and trust in politicians. As the theory on human values would predict-conceptualising values as stable, deep-seated cognitive traits that are likely to be distal from political attitudes and even more so from political behaviour-their effects on voting at the 2016 referendum were thus largely Table 2 Standardised probit coefficients of ten basic human values predicting voting to remain in the EU (bivariate models).

Table 3
Standardised probit coefficients of European identity, opposition to immigration and trust in politicians predicting voting to remain in the EU (bivariate models).

Predictor of voting to remain Coefficient
European identity (n = 1,406) 0.523*** (0.027) Opposition to immigration (n = 1,951) −0.573*** (0.029) Trust in politicians (n = 1,408) 0.162*** (0.037) Notes: standard errors in parentheses; ***p < 0.001, **p < 0.01, *p < 0.05.  via the more proximal attitudes. We found empirical evidence to support this framework, particularly regarding the value-based antecedents of European identity and opposition to immigration. We also found that the latter in turn affected voting in the referendum. Furthermore, the total effects of values on voting to remain in the EU in the referendum was approximately as strong as the direct effects of the more proximal predictors, European identity and opposition to immigration. There was no empirical evidence, though, for the effect of trust in politicians on voting to remain after we controlled for sociodemographic variables and the other values. It could well be the case that including conservation values, education, and subjective income in the model, all of which had a significant correlation with trust in politicians, reduced the remaining explanatory power of the variable trust. Alternatively, the notion that voting by Britons on EU membership was a proxy for expressing levels of antipolitician and populistic attitudes may have been overstated and, instead, Britons voted according to their levels of Europeanness and attitudes to immigration, both being driven by the more deep-seated values.
The ultimate reasons why conservation and universalist values influenced vote choice in the 2016 referendum relate in large part to popular perceptions of the EU and the UK's relationship with it. Some of these perceptions, and thus relationships, are likely to be common across Europe, both in and out of the EU. Those who value universalism are likely to support supranational bodies given the additional value that those individuals place on understanding, appreciation, tolerance, and protection for the welfare of all people. On the other hand, the negative relationship between valuing conservation and support for membership may be more UK-specific, given Britain's long-time characterisation as an "awkward partner" of "reluctant Europeans" who are "European but not European enough" (e.g. Carl et al., 2019). Indeed, Britons have for a long-time held the weakest sense of common European identity of any member state (ibid), resulting from a, by now, many-times repeated list of political, social and economic divergences with the core member states, in terms of the original six or the Eurozone. All this means that, probably particularly for Britons, the EU is more likely to be seen as a threat to security, tradition and societal conformity. In other countries, there may be no such connotations, or even the reverse, so that the EU is seen as a positive bulwark of these three values (plausibly the case in member states that face an alternative security threat, such as the Baltics, internal instability or far-left political agitation, or whose national political traditions are shorter lived than the EU), reflecting the important national-context element of attitudes to EU membership (De Vries, 2018). Future research should explore the relationship between values and attitudes to the EU outside of the UK, keeping these historic trajectories in mind.
Notwithstanding these findings, our study is not without limitations. First, our behavioural measure is retrospective. Although data collection took place relatively shortly after the referendum, people might have forgotten how they actually voted, or they might have concealed their true voting behaviour due to social desirability. The use of survey data has the advantage of accessing a large database on values, attitudes and behaviour of UK respondents, but does not allow us to observe more directly actual behaviour. Second, although values are considered more abstract and general than attitudes such as European identity or opposition to immigration, and consequently are more likely to affect such attitudes rather than be influenced by them, we cannot exclude the possibility that the causal chain in our model also operated in the other direction, that is, from attitudes (or behaviour) to the values. However, we consider this relatively unlikely, given the personal rather than political nature of Schwartz's values and their greater stability over time than political attitudes, which are far more context-specific (both across space and time) and thus more volatile. Future studies with experimental designs or panel data would shed more light on the causal directions and mechanisms between values, attitudes and voting behaviour.
In sum, the study suggests that basic human values matter in the explanation of the decision of UK respondents to remain in the EU. They were more important than socio-demographic characteristics and similarly important to more proximal predictors like European identification or opposition to immigration: European identity was likely to result in a vote to remain while opposition to immigration was likely to be associated with a vote to leave the EU; universalist individuals were more and conservative individuals less likely to vote in favour of remaining in the EU. Notes: standard errors in parentheses; ***p < 0.001, **p < 0.01, *p < 0.05.